SIS code: 
Jan Cuřín

NPFL123 – Dialogue Systems


This course is a detailed introduction into the architecture of spoken dialogue systems, voice assistants and conversational systems (chatbots). We will introduce the main components of dialogue systems (speech recognition, language understanding, dialogue management, language generation and speech synthesis) and show alternative approaches to their implementation.

The lab sessions will be dedicated to implementing a simple dialogue system and selected components (via weekly homework assignments). We will use Python and a version of our Dialmonkey framework for this.

Logistics (spring 2023)


The course will be taught in English, but we're happy to explain in Czech, too.

Time & Place

In-person lectures and labs take place in the Malá Strana building.

  • Lectures: Mon 15:40, room S9 (1st floor)
  • Labs: Mon 17:20, room S9 (1st floor)

In addition, we plan to stream both lectures and lab instruction over Zoom and make the recordings available on Youtube (under a private link, on request). We'll do our best to provide a useful experience.

  • Zoom meeting ID: 981 1966 7454
  • Password is the SIS code of this course (capitalized)

There's also a Slack workspace you can use to discuss assignments and get news about the course. Invite links will be sent out to all enrolled students by the start of the semester. Please contact us by email if you want to join and haven't got an invite yet.

Passing the course

To pass this course, you will need to take an exam and do lab homeworks. There's a 60% points minimum for the exam and 50% for the homeworks to pass the course. See more details here.

Topics covered

Dialogue systems schema
  • Dialogue system types & formats (open/closed domain, task/chat-oriented)
  • What happens in a dialogue (linguistic background)
  • Dialogue system components
    • speech recognition
    • language understanding, dialogue state tracking
    • dialogue management
    • language generation
    • speech synthesis
  • Dialogue authoring tools (IBM Watson Assistant/Google Assistant/Amazon Alexa)
  • Voice assistants & question answering
  • Chatbots
  • Data for dialogue systems
  • Dialogue systems evaluation


PDFs with lecture slides will appear here shortly before each lecture (more details on each lecture are on a separate tab). You can also check out last year's lecture slides.

1. Introduction Slides Domain selection Questions

2. Data & Evaluation Slides Dataset exploration Questions

3. What happens in a dialogue? Slides Rule-based Natural Language Understanding Questions

4. Natural Language Understanding Slides Statistical Natural Language Understanding Questions

5. Neural NLU + State Tracking Slides Belief State Tracking Questions

6. Dialogue Policy (non-neural) Slides Dialogue Policy Questions

7. Neural policies & Natural Language Generation Slides API/Backend Calls Questions

8. Voice Assistants & Question Answering Slides Template NLG Questions

9. Dialogue Tooling Slides Service Integration Questions

10. Speech Synthesis Slides Grapheme-to-phoneme conversion Questions

11. Speech Recognition Slides Digits ASR Questions

12. Chatbots Slides Retrieval chatbot Questions


A list of recommended literature is on a separate tab.


1. Introduction

 13 February Slides Domain selection Questions

  • What are dialogue systems
  • Common usage areas
  • Task-oriented vs. non-task oriented systems
  • Closed domain, multi-domain, open domain
  • System vs. user initiative in dialogue
  • Standard dialogue systems components

2. Data & Evaluation

 20 February Slides Dataset exploration Questions

  • How to get data for building dialogue systems
  • Available corpora/datasets
  • Annotation
  • Data splits
  • Evaluation metrics -- subjective & objective, intrinsic & extrinsic
  • Significance checks

3. What happens in a dialogue?

 27 February Slides Rule-based Natural Language Understanding Questions

  • Dialogue turns
  • Utterances as acts, pragmatics
  • Grounding, grounding signals
  • Deixis
  • Conversational maxims
  • Prediction and adaptation

4. Natural Language Understanding

 7 March Slides Statistical Natural Language Understanding Questions

  • What needs to be handled to understand the user
  • How to represent meaning: grammars, frames, graphs, dialogue acts (“shallow parsing”)
  • Rule-based NLU
  • Classification-based NLU (features, logistic regression, SVM)
  • Sequence tagging (HMM, MEMM, CRF)
  • Handling speech recognition noise

5. Neural NLU + State Tracking

 13 March Slides Belief State Tracking Questions

  • Some basics about neural networks
  • How to use neural networks for NLU: neural classifiers and sequence taggers
  • What is dialogue state, what is belief state, and what they're good for
  • Dialogue as a Markov decision process (MDP)
  • Dialogue trackers: generative and discriminative
  • Static and dynamic trackers

6. Dialogue Policy (non-neural)

 20 March Slides Dialogue Policy Questions

  • What's a dialogue policy -- how to choose the next action
  • Finite-state, frame-based, and rule-based policies
  • Reinforcement learning basics
  • Value and policy optimization (SARSA, Q-learning, REINFORCE)
  • Mapping to POMPDs (partially observable MDPs)
  • Summary space (making it tractable)
  • User simulation

7. Neural policies & Natural Language Generation

 27 March Slides API/Backend Calls Questions

  • Deep reinforcement learning
  • Deep Q-Networks, Policy Networks
  • Natural language generation
  • Sentence planning & Surface realization
  • Templates
  • Rule-based approaches
  • Neural: seq2seq, RNNs, Transformers

8. Voice Assistants & Question Answering

 3 April Slides Template NLG Questions

  • What are voice assistants
  • Where, how and how much are they used
  • What are their features and limitations
  • What is question answering
  • Basic question answering techniques
  • Knowledge graphs

9. Dialogue Tooling

 17 April Slides Service Integration Questions

  • What are the standard tools for building dialogue systems on various platforms
  • IBM Watson, Google Dialogflow, Alexa Skills Kit
  • How to define intents, slots, and values
  • How to build your own basic dialogues

10. Speech Synthesis

 24 April Slides Grapheme-to-phoneme conversion Questions

  • Human articulation
  • Phones, phonemes, consonants, vowels
  • Spectrum, F0, formants
  • Stress and prosody
  • Standard TTS pipeline
  • Segmentation
  • Grapheme-to-phoneme conversion
  • Formant-based, concatenative, HMM parametric synthesis
  • Neural synthesis

11. Speech Recognition

 15 May Slides Digits ASR Questions

  • Basics of how speech recognition works
  • Main pipeline: speech activity detection, preprocessing, acoustic model, decoder
  • Features -- MFCCs
  • Acoustic model with neural nets
  • Decoding -- language model
  • End-to-end speech recognition

12. Chatbots

 22 May Slides Retrieval chatbot Questions

  • Non-task-oriented systems and their specifics
  • rule-based, retrieval, generative, hybrid approaches
  • Turing test, Alexa Prize

Homework Assignments

There will be 12 homework assignments, each for a maximum of 10 points. Please see details on grading and deadlines on a separate tab. Note that there's a 50% minimum requirement to pass the course.

Assignments should be submitted via Git – see instructions on a separate tab. Please take special care about naming your Git branches and files the way they're given in the assignments. If our automatic checks don't find your files, you'll lose points!

You can run our automatic checks from your checkout -- have a look at Note that you need to update your checkout to have the necessary files -- and you will need to update it a few days before each assignment's due date since we'll be adding checks on the go.

All deadlines are 23:59:59 CET/CEST.


1. Domain selection

2. Dataset exploration

3. Rule-based Natural Language Understanding

4. Statistical Natural Language Understanding

5. Belief State Tracking

6. Dialogue Policy

7. API/Backend Calls

8. Template NLG

9. Service Integration

10. Grapheme-to-phoneme conversion

11. Digits ASR

12. Retrieval chatbot

1. Domain selection

 Presented: 13 February, Deadline: 2 March

You will be building a dialogue system in (some of) the homeworks for this course. Your first task is to choose a domain and imagine how your system will look like and work like. Since you might later find that you don't like the domain, you are now required to pick two, so you have more/better ideas later and can choose only one of them for building the system.


The required steps for this homework are:

  1. Pick two domains of your liking that are suitable for building a dialogue system. Think of a reasonable backend (see below).

  2. Write 5 example system-user dialogues for both domains, which are at least 5+5 turns long (5 sentences for both user and system). This will make sure that your domain is interesting enough. You do not necessarily have to use English here (but it's easier if we understand the language you're using -- ask us if unsure, Czech & Slovak are perfectly fine).

  3. Create a flowchart for your two domains, with labels such as “ask about phone number”, “reply with phone number”, “something else” etc. It should cover all of your example dialogues. Feel free to draw this by hand and take a photo, as long as it's legible.

    • It's OK if your example dialogues don't go all in a straight line (e.g. some of them might loop or go back to the start).

Files to commit

Please stick to the file naming conventions -- you will lose points if you don't!

  • hw1/ with short commentary on both domains (ca. 10-15 sentences) -- what they are, what features you'd like to include, what will be the backend.

  • hw1/examples-<domain1>.txt, hw1/examples-<domain2>.txt -- 5 example dialogues for each of the domains (as described above). Use a short domain name, best with just letters and underscores.

  • hw1/flowchart-<domain1>.{pdf,jpg,png}, hw1/flowchart-<domain2>.{pdf,jpg,png} -- the flowcharts for each of the domains, as described above.

See the instructions on submission via Git -- create a branch and a merge request with your changes. Make sure to name your branch hw1 so we can find it easily.


You may choose any domain you like, be it tourist information, information about culture events/traffic, news, scheduling/agenda, task completion etc. You can take inspiration from stuff presented in the first lecture, or you may choose your own topic.

Since your domain will likely need to be connected to some backend database, you might want to make use of some external public APIs -- feel free to choose under one of these links:

You can of course choose anything else you like as your backend, e.g. portions of Wikidata/DBPedia or other world knowledge DBs, or even a handwritten “toy” database of a meaningful size, which you'll need to write to be able to test your system.

2. Dataset exploration

 Presented: 20 February, Deadline: 9 March

The task in this lab is to explore dialogue datasets and find out more about them. Your job will thus be to write a script that computes some basic statistics about datasets, and then try to interpret the script's results.


  1. Take a look at the Dialog bAbI Tasks Data 1-6 dataset. Read the description of the data format in the repo and on the original website. You'll be working with Tasks 5 and 6 (containing full generated dialogues and DSTC2 data). Use the training sets for Task 5 and Task 6.

  2. Write a script that will read all turns in the data and separate the user and system utterances in the training set.

    • Make the script ignore any search results lines in the data (they don't contain a tab character).
    • If the script finds a turn where the user is silent (the user turn contains only <SILENCE>), it should concatenate the system response from this turn to the previous turn. Note that this may happen on multiple consecutive turns, and the script should join all of these together into one system response.
      • If <SILENCE> is the first word in the dialogue, just delete it.
    • For tokenization, you should use the word_tokenize function from the nltk package.
  3. Implement a routine that will compute the following statistics for both bAbI tasks for system and user turns (separately, i.e., 4 sets of statistics altogether):

    • data length (total number of dialogues, turns, words)
    • mean and standard deviations for individual dialogue lengths (number of turns in a dialogue, number of words in a turn)
    • vocabulary size
    • Shannon entropy and bigram conditional entropy, i.e. entropy conditioned on 1 preceding word (see lecture 3 slides)

    Commit this file as hw2/

  4. Along with your script, submit also a dump of the results. The results should be formatted as JSON file with the following structure:

        "dialogues_total": XXX,
        "turns_total": XXX,
        "words_total": XXX,
        "mean_dialogue_turns": XXX,
        "stddev_dialogue_turns": XXX,
        "mean_dialogue_words_per_turn": XXX,
        "stddev_dialogue_words_per_turn": XXX,
        "vocab_size": XXX,
        "entropy": XXX
        "cond_entropy": XXX
    "task5_system": ...
    "task6_user": ...
    "task6_system": ...

(Create a dict and use json.dump for this.) Commit the json file as hw2/results.json.

  1. Add your own comments, comparing the results between the two bAbI Tasks. 3-5 sentences is enough, but try to explain why you think the vocabulary and entropy numbers are different.

    Put your comments in Markdown as hw2/

Files to commit

There are empty files ready for you in the repo the right places, you just need to fill them with information.

Just to sum up, the files are:

  • hw2/ -- your data analysis script
  • hw2/results.json -- JSON results of the analysis
  • hw2/ -- your comments

Create a branch and a merge request containing (changes to) all requested files. Please keep the filenames and directory structure.


  • Don't worry too much about the exact numbers you get, slight variations in implementation may cause them to change. We won't penalize it if you don't get the exact same numbers as us, the main point is that your implementation should be reasonable (and you shouldn't be off by orders of magnitude).
  • Don't worry about system api_calls, just treat them as a normal system turn.

3. Rule-based Natural Language Understanding

 Presented: 27 February, Deadline: 16 March

In this assignment, you will design a dialogue system component for Natural Language Understanding in a specific domain. For completing it, you will use our prepared Dialmonkey dialogue framework (which is the base of your Gitlab repository), so you can test the outcome directly.


  1. Make yourself familiar with the Dialmonkey-npfl123 repository you cloned for the homeworks. Read the README and look around a bit to get familiar with the code. Have a look at the 101 Jupyter notebook to see some examples.

  2. Recall the domains you picked in the first homework assignment and choose one of them. If you've changed your mind in the meantime, you can even pick a different domain.

  3. Think of the set of dialogue acts suitable to describe this domain, i.e., list all intents, slots and values that will be needed (some slots may have open sets of values, e.g. “restaurant name”, “artist name”, “address” etc.). List them, give a description and examples in Markdown under hw3/

  4. Create a component in the dialogue system (as Python code) that:

    • inherits from dialmonkey.component.Component
    • is placed under dialmonkey.nlu.rule_<your-domain>
    • implements a rule-based NLU for your domain -- i.e., given user utterance, finds its intent, slots and values
    • yields Dialogue Act items you designed in step 3 (as DA objects).

    Please only use the core Python libraries and those listed in requirements.txt. If you have a good reason to use a different library from PyPi, let us know and we can discuss adding it into the requirements (but this will be global for everyone).

  5. Create a YAML config file for your domain in the conf directory. You can use the sample_conf.yaml file or nlu_test.yaml as a starting point. These files are almost identical, just have a look at the I/O setup if you're interested. Note that instead of a policy and NLG components, a system with this config will simply reply with the NLU result.

  6. Write at least 15 distinct utterances that demonstrate the functionality of your class (a tab-separated file with input + corresponding NLU result, one-per-line). Make sure your NLU gives you the same results. The test utterances can (but don't have to) be taken over from the example dialogues you wrote earlier for your domain. Save them as hw3/examples.tsv.

Files to include in your merge request

If you update your dialmonkey-npfl123 checkout, there will be empty files ready for you in the right place, just rename according to your domain (e.g. change <your_domain> to restaurant, bus etc.) and fill them with the required content.

  • Lists of intents, slots and values in hw3/
  • Your NLU component in dialmonkey/nlu/rule_<your_domain>.py
  • Your configuration file in conf/nlu_<your_domain>.yaml
  • Example test utterances & outputs in hw3/examples.tsv

Create a branch and a merge request containing (changes to) all requested files. Please keep the filenames and directory structure.


Use regular expressions or keyword matching to find the intents and slot values (based on the value, you'll know which slot it belongs to).

If you haven't ever used regular expressions, have a look some tutorials:

Note that you might later need to improve your NLU to handle contextual requests, but you don't need to worry about this now. For instance, the system may ask What time do you want to leave? and the user replies just 7pm. From just 7pm (without the preceding question), you don't know if that's a departure or arrival time. Once you have your dialogue policy ready and know how the system questions look like (which will be the 6th homework), you'll be able to look at the last system question and disambiguate. For now, you can keep these queries ambiguous (e.g. just mark the slot as “time”).

4. Statistical Natural Language Understanding

 Presented: 6 March, Deadline: 23 March

In this assignment, you will build and evaluate a statistical Natural Language Understanding component on the DSTC2 restaurant information data. For completing it, you will use the Dialmonkey framework in your code checkout so you can test the outcome directly.


  1. Locate the data in your Dialmonkey-NPFL123 repository in data/hw4/. This is what we'll work with for this assignment.

  2. Implement a script that trains a statistical model to predict DAs. It shouldn't predict the whole DA as a single classifier, rather it should classify the correct value for each intent-slot pair where applicable (e.g. inform(food) has multiple possible values) and classify a binary 0-1 for each intent-slot pair that can't have different values (e.g. request(price) or bye() ).

    Don't forget that for the multi-value slots, you'll need a “null” value too.

    You can use any kind of statistical classifier you like (e.g. logistic regression, SVM, neural network), with any library of your choice (e.g. Scikit-Learn, Tensorflow, Pytorch ).

    Note that we're not doing slot tagging since the words in the texts aren't tagged with slot values.

  3. Train this model on the training set. You can use the development set for parameter tuning. Using dialmonkey.DA.parse_cambridge_da() should help you get the desired DA values out of the textual representation. Do not look at the test set at this point!

  4. Evaluate your model on the test set and report the overall precision, recall and F1 over dialogue act items (triples of intent-slot-value).

    Use the script provided in dialmonkey.evaluation.eval_nlu. You can run it directly from the console like this:

    ./dialmonkey/evaluation/ -r data/hw4/dstc2-nlu-test.json -p hw4/predicted.txt

    The script expects reference JSON in the same format as your data here, and a system output with one DA per line. You can have a look at conf/nlu_test.yaml to see how to get one-per-line DA output.

    For the purpose of our evaluation script F1 computation, non-null values count as positives, null values count as negatives. Whether they're true or false depends on whether they're correctly predicted.

  5. Implement a module in Dialmonkey that will load your NLU model and work with inputs in the restaurant domain. Create a copy of the nlu_test.yaml config file to work with your new NLU.

Files to include in your merge request

  • Your statistical NLU module as dialmonkey/nlu/ and your trained model. The filename for the model will depend on your implementation, it just needs to be loaded automatically for your model to work. Preferrably store in the same directory, but it could be anywhere else if needed. If the file is too big for Git (>10MB), share it using a cloud service (e.g. CESNET OwnCloud) and make your NLU module download it automatically in __init__().
  • Your YAML config file under conf/nlu_dstc.yaml.
  • Your training script as hw4/
  • Your predicted output as hw4/predicted.txt and a short evaluation report under hw4/ (including your F1 scores).

Important implementation notes

Note 1: Please do not use any Python libraries other than the ones in requirements.txt, plus the following ones: torch (Pytorch), tensorflow, pytorch-lightning, torchtext, transformers (Huggingface). Note that scikit-learn is included already. If you need any others, please let us know beforehand.

Note 2: Please make sure that your code doesn't take more than a minute to load + classify the first 20 entries in the test data. For the sake of model storage, it's better to choose a smaller one, especially if you choose to play with pretrained language models.

Note 3: And this one is for all further assignments -- Do not use absolute file paths in your solution! You may want to try out os.path.dirname(__file__), which gets you the directory of the current source file. If you need to use slashes in a (relative!) path, use either os.path.join instead, or use forward slashes and a pathlib.Path object (simply create it using filename = Path(filename)). This will handle the slashes properly both on Windows and Linux. Note that we're mainly checking on Linux, so any backslashes will break our workflow and we may deduce points.


  • Start playing with the classifier separately, only integrate it into Dialmonkey after you've trained a model and can load it.

  • If you have never used a machine learning tool, have a look at the Scikit-Learn tutorial. It contains most of what you'll need to finish this exercise.

  • You'll need to convert your texts into something your classifier understands (i.e., some input numerical features). You can probably do very well with just “bag-of-words” as input features to the classifier -- that means that you'll have a binary indicator for each word from the training data (e.g. word “restaurant”). The feature for the word “restaurant” will be 1 if the word “restaurant” appears in the sentence, 0 if it doesn't. You can also try using the same type of features for bigrams. Have a look at the DictVectorizer class in Scikit-Learn.

  • For Scikit-Learn, you can use pickle to store your trained models.

  • To easily load JSON files, you can use the SimpleJSONInput class.

  • You better don't use the naive Bayes classifier, it doesn't work well on this data – basically anything else works better (you won't lose any points if you use naive Bayes, just don't expect good performance).

5. Belief State Tracking

 Presented: 13 March, Deadline: 30 March

This week, you will build a simple probabilistic dialogue/belief state tracker to work with NLU for both your own domain of choice (3rd assignment) and DSTC2 (4th assignment).


  1. Implement a dialogue state tracker that works with the dial.state structure (it's a dict) and fills it with a probability distribution of values over each slot (assume slots are independent), updated during the dialogue. Don't forget None is a valid value, meaning “we don't know/user didn't say anything about this”.

    At the beginning of the dialogue, each slot should be initialized with the distribution {None: 1.0}.

    The update rule for a slot, say food, should go like this:

    • Take all mentions of food in the current NLU, with their probabilities. Say you got Chinese with a probability of 0.7 and Italian with a probability of 0.2. This means None has a probability of 0.1.
    • Use the probability of None to multiply current values with it (e.g. if the distribution was {'Chinese': 0.2, None: 0.8}, it should be changed to {'Chinese': 0.02, None: 0.08}.
    • Now add the non-null values with their respective probabilities from the NLU. This should result in {'Chinese': 0.72, 'Italian': 0.2, None: 0.08}.

    Note that you should be using probability estimates in your statistical NLU from the 4th assignment) now (for your rule-based NLU on own domain from the 3rd assignment), just assume a probability of 1 if your patterns match).

    Set a threshold for low-probability NLU outputs that you'll ignore in the tracker (good choice is 0.01-0.05), so that the outputs are not too messy.

  2. Create new configuration files both for your own rule-based domain and for DSTC2, which include NLU and the tracker. In addition, use dialmonkey.policy.dummy.ReplyWithState as the output (so the system's response is now the current dialogue state).

    You can use conf/dst_test.yaml as a starting point (just put in your NLU and tracker).

  3. Run your NLU + tracker over your NLU examples from 3rd assignment and the first 20 lines from the DSTC2 development data (file data/hw4/dstc2-nlu-dev.json ) from the 4th assignment data and save the outputs to a text file for both.

    Treat the examples as a single dialogue, even though such a “dialogue” doesn't make any logical sense. Your tracker should be able to handle it.


  • You can have a look at an example dialogue with commentary for the update rule in a separate file (it's basically the same stuff as above, just more detailed).

  • Remember that all this is just about slots, not intents – do not consider intents as part of the state. You generally don't need to track past intents, just the current intent from NLU is enough to guide the system.

  • If your statistical NLU is built as classifiers for intent-slot pairs (as foreseen by default in HW4), you can just use the probabilities coming from the inform intent for the given slot. The best way would be to add up "positive" intents such as inform or confirm and discount negative ones, such as deny, but we don't require you to do this as part of the assignment.

Files to include in your merge request

  • Your tracker code under dialmonkey/dst/
  • Your updated configuration files for both domains under conf/dst_<my-domain>.yaml and conf/dst_dstc.yaml.
  • Your text files with the outputs into hw5/outputs_<my-domain>.txt and hw5/outputs_dstc.txt.

6. Dialogue Policy

 Presented: 20 March, Deadline: 13 April (with HW7!)

This week, you will build a rule-based policy for your domain.


  1. Implement a rule-based policy that uses the current NLU intent (or intents, coming from your NLU) and the dialogue state (coming from your tracker) to produce system action dialogue acts. You can use just the most probable value for each slot, assuming its probability is higher than a given threshold (e.g. 0.7). Since your NLU is rule-based, the actual value of the threshold doesn't matter much (should be >0): you'll likely have zero or one value with a probability of 1, the rest with a probability zero.

    The policy should:

    • Check the current intent(s) and split the action according to that
    • Given the intent, check that the state contains all necessary slots to respond:
      • if it does, fill in a response system DA into the dial.action field.
      • if it doesn't, fill in a system DA requesting more information into the dial.action field.

    You can use the flowcharts you built in HW1 to guide the policy's decisions (but it's fine if you don't stick to them).

    For now, skip any API queries and hardcode the responses (you will build the actual backend a week later; or you can wait and build the two assignments together).

  2. Save the policy under dialmonkey.policy.rule_<your_domain> and create a configuration file for your own domain that will include it.

  3. In your new configuration file, include dialmonkey.policy.dummy.ReplyWithSystemAction as the last step -- this replaces dialmonkey.policy.dummy.ReplyWithState you used previously (don't worry that you have two things from the policy package in your pipeline).

  4. Check that your policy returns reasonable system actions for each of your NLU test utterances, if you treat each utterance as a start of a separate new dialogue (it's OK if it's not exactly what is meant, the utterances will be taken out of context).

    Run your policy over your NLU test utterances, each taken as a start of a separate dialogue, and save the outputs to a text file. You can either do this by hand, or you may consider creating a script that creates an instance of ConversationHandler and uses handler.run_dialogue() -- see the [](101 Jupyter notebook). Outputs can be saved to a file using dialmonkey.output.text.FileOutput.

Files to include in your repository

  • Your policy (dialmonkey/policy/rule_<your_domain>.py) and your updated configuration file (conf/act_<your_domain>.yaml).
  • The outputs of your policy on your NLU test utterances as hw6/outputs.txt.

Note: If you're submitting together with HW7, make just a single merge request, with a branch labelled hw7, but include all the files required for both assignments.

Note 2: To make testing smooth, please merge your HW3 merge request into master & merge that into your submission MR for HW6. This will make your NLU examples file available for us, so we see not just the output but also the corresponding inputs in the repository.


  • Your policy will most probably be a long bunch of if-then-else statements. Don't worry about it. You may want to structure the file a bit so it's not a huge long function though -- e.g., add the handling into separate functions, if it's not just 1 line.

  • You may want to complete this homework together with the next one, which will be about backend integration. That's why the deadline is in 3 weeks, not 2.

7. API/Backend Calls

 Presented: 27 March, Deadline: 13 April (with HW6)

This week is basically a continuation of the last one -- filling in the blank API call placeholders you created last time. If you want to, you can complete the 6th and 7th homework at the same time (the deadline is the same).


  1. Implement all API/backend queries you need for the policy you implemented in the last homework assignment. The implementation can be directly inside dialmonkey.policy.rule_<your_domain>, or you can create a sub-package (a subdirectory, where you put the main policy inside and any auxiliary stuff into other files in the same directory).

    It depends on your domain if the implementation will be external API queries (preferrable) or some kind of internal database (e.g. a CSV table). In case you implement a local database, please add at least 10 different entries (let us know if you think this doesn't make sense for your domain).

  2. Test your policy with outputs on at least 3 of the test dialogues you created in the 1st homework assignment. You can of course make alterations if your policy doesn't behave exactly as you imagined the first time. If you changed your domain, you can draft 3 dialogues of 5+ turns.

    Also, don't worry that the output is just dialogue acts at the moment.

Files to include in your repository

  • Commit your policy (dialmonkey/policy/rule_<your_domain>.py), updated with API calls.
  • Put logs of your test dialogues as hw7/outputs.txt (with user inputs and system output acts).

Note: if you're submitting together with HW6, make just a single merge request, with a branch labelled hw7, but include all the files required for both assignments.

Implementation notes

For hw7/outputs.txt, please make the format like this:

U: Hello, how are you?
S: hello()&request(area)
U: Thank you, goodbye!
S: bye()

U: Hello, I need a cheap restaurant
S: hello()&confirm(price=cheap)&request(area)

Make all lines start with U: and S: marking the user and system turns, with system turns consisting of dialogue acts (in Cambridge or Dialmonkey format). Put an empty line between different dialogues.

Update: There's now a class called DialogueLogOutput, which will sort out the format for you. Just use it as your output stream (see the README on how to do that). Note that you'll need to update from upstream to be able to use that class in your code!


  • If you haven't accessed an external API using Python, check out the requests library. It makes it easy to call external APIs using JSON. You can get the result with just a few lines of code.

8. Template NLG

 Presented: 3 April, Deadline: 20 April

In this homework assignment, you will complete the text-based dialogue system for your domain by creating a template-based NLG component.


  1. Implement a template-based NLG with the following features:

    • The NLG system is (mostly) generic and can load templates for your domain from a JSON or YAML file (only one of these formats is fine!), showing a DA -> template mapping.

    • The NLG system is able to prioritize a mapping for a specific value -- e.g. inform(price=cheap) -> “You'll save money here.” should get priority over inform(price={price}) -> “This place is {price}.”

    • The NLG system is able to put together partial templates (by concatenating), so you can get a result for e.g. inform(price=cheap,rating=3) even if you only have templates defined for inform(price={price}) and inform(rating={rating}), not the specific slot combination. So if you have templates inform(price={price}) -> “The place is {price}.” and inform(rating={rating}) -> “The place is rated {rating} stars.”, your output for inform(price=cheap,rating=3) should be “The place is cheap. The place is rated 3 stars.”

      This doesn't need to search for best coverage, just take anything that fits, such as templates for single slots if you don't find the correct combination.

    • The system is able to produce multiple variations for certain outputs, e.g. bye() -> Goodbye. or Thanks, bye!

  2. Create templates that cover your domain well.

  3. Save your NLG system under dialmonkey.nlg.templates_<your-domain> and add it into your conf/text_<your-domain>.yaml configuration file.

  4. Test your NLG system with the test dialogues you used in the previous assignment.

Implementation instructions

Make sure your template NLG is really domain-general. We're going to test it on a different domain from yours! You can have specific additions for your domain in there, just make sure it'll work reasonably well with any slot names.

Access to template file: In your NLG class, include a constructor method with this signature: def __init__(self, config). Call super().__init__(config) as the first thing inside this constructor method. Include a def load_templates(self, filename) function which loads your templates and call it in the constructor (note this method will be called in our tests!).

You can pass the filename to the constructor (inside config) via the YAML config file, e.g. you add:

    - dialmonkey.nlg.templates_<your-domain>:
        templates_file: dialmonkey/nlg/templates_<your-domain>.<yaml|json>

And you can use config['templates_file'] in the constructor to get the file path.

Paths warning (repeat): If you need slashes in the path, use forward slashes and a pathlib.Path object (simply create it using filename = Path(filename)), it will handle the slashes properly both on Windows and Linux. Do not use absolute paths! You may want to try out os.path.dirname(__file__), which gets you the directory of the current source file.

Templates format: Use a dict where keys are DAs (either in native triple style, or Cambridge style) and values are lists of templates (variants), with slot placeholders marked using curly braces. Example YAML:

- This place is cheap.

- This place is {price}.

- Goodbye.
- Thanks, bye!

Example JSON:

"inform(price=cheap)": [
    "This place is cheap."

"inform(price={price})": [
    "This place is {price}."

"bye()": [
    "Thanks, bye!"

Setting the reply: Call dial.set_system_response(text) where text is your NLG output, so the system actually uses your NLG to respond. Remove ReplyWithSystemAction (used in HW6 & HW7) from your config file, it's no longer needed :-).

Files to include in your merge request

  • Your NLG implementation in dialmonkey/nlg/templates_<your-domain>.py
  • Your templates file in dialmonkey/nlg/templates_<your-domain>.<yaml|json>
  • A full configuration file for your domain, which includes the NLG system, under conf/text_<your-domain>.yaml
  • Logs of the test dialogues from HW7, now with NLG output, in hw8/outputs.txt

9. Service Integration

 Presented: 24 April, Deadline: May 11

In this homework, you will integrate the chatbot for your domain into an online assistant of your choice (Google/Alexa/Facebook/Telegram).


  1. Choose a service that you want to use for this homework. We prepared some instructions for Google Dialogflow, Alexa Skills, Facebook Messenger, and Telegram. but can also use the IBM Watson Assistant as shown in the lecture by Honza. Note that Dialogflow and Alexa are unfortunately not available for Czech.

  2. Implement the frontend on your selected platform. You can either carry over intents, slots & values from your NLU directly into Dialogflow/Alexa/Watson, or you can work with free text and run the NLU in the backend. For Messenger and Telegram, that's the only option (but their frontend basically comes for free).

  3. Implement a backend that will connect to your frontend – handle its calls and route them to your dialogue system in Dialmonkey (either with NLU already processed, or with free text as input). You can use the get_response method in dialmonkey.conversation_handler.ConversationHandler to control the system outside of the console. Don't forget to save context in between requests. However, you can assume for simplicity that the bot will always have just one dialogue, i.e. you do not have to care about parallel conversations.

  4. Link your frontend to your backend (see Hints below).

Detailed instructions

Amazon Alexa
  • Alexa allows you to run NLU directly in your backend but it's a bit tricky -- the only way to get free text is to use the SearchQuery built-in slot. You can set up an intent where the only part of the utterance is this slot.

  • For implementing backend, you can use the Flask-Ask package as a base. You can have a look at an Alexa Skill Ondrej made for inspiration (not many docs, though, sorry).

  • Set your backend address under “Endpoint” (left menu).

Google Dialogflow
  • To make use of you NLU, you can make use of the Default fallback intent which gets triggered whenever no other intent is recognized.
  • For backend implementation, you can use Flask-Assistant.
  • Set your backend address under “Webhooks” for the individual intents (under the “Fulfillment” menu of each intent). If you want to get free text of the requests, have a look at this snippet.
Facebook Messenger

For Messenger, you need to perform several steps, however, it allows you to work with textual inputs directly. In general, you can follow the tutorial. Here are the important steps you need to complete:

  • Implement a webserver using Flask. It might be a good idea to start from the example implementation (feel free to reuse it).
    See the tutorial here for other options.
  • You need to implement GET method handler for verification and POST method handler to receive message and send the reply. You can also use pymessenger, though the code isn't maintained.
  • Set up your webserver on a public URL.
  • The verification step requires two tokens -- a verification token (an arbitrary string that you choose) and an access token (which you'll get from Facebook).
  • Visit the dev page. Create an account, add a Messenger app and create a sample Facebook page (in the “App page” category).
  • Link the page to your app (under “Messenger” – “Settings”) and obtain the access token for use in your webserver (generate & save into your code).
  • Add a callback URL pointing to your webserver and the verification token (the one you chose).

Telegram also allows direct text input. There's also a handy Python-Telegram-Bot library that has a webserver built in.

  • You need to get a telegram bot API token from the BotFather -- it's pretty straightforward, you talk to this “one bot to rule them all”, ask it for /newbot and it'll guide you through the process. You can find more info on their documentation page.
  • With the use of your API token, you need to create the telegram bot using Python-Telegram-Bot. These examples should be a good way of starting it. There's even a tutorial. In general, you'll need to implement a command handler that'll pass on the message from Telegram to your system. No web server is necessary, Python-Telegram-Bot has one built in.
  • Start up your bot and you can talk to it on Telegram.

Files to include in your merge request

  • Commit your frontend export into hw9/:
    • In Alexa, go to “JSON Editor” in the left menu and copy out the contents into a file intent_schema.json.
    • In Dialogflow, go to your agent settings (cogged wheel next your agent/skill/app name on the top left), then select the “Export & Import” tab and choose “Export as ZIP”. Please commit the resulting subdirectory structure, not the ZIP file.
    • Nothing is required for Messenger and Telegram at this point.
  • Commit your bot server code into hw9/ This code will probably import a lot from Dialmonkey and require it to be installed -- that's expected.
    • Make sure that python hw9/ will run the bot with the default settings. Use if __name__ == '__main__' for that, so we can import your code for tests.
    • Include a function called reply_<google|alexa|facebook|telegram> that replies to a message (depending on the service) or calls a function that does that (passing on whatever parameters are needed on the given platform). This corresponds to:
      • Google: A Flask-Assistant function with the decorator @assist.action('Default fallback intent')
      • Alexa: A Flask-Ask function with the decorator @ask.intent('YourDefaultIntent')
      • Facebook: The webhook function with the Flask decorator @app.route('/', methods=['POST'])
      • Telegram: A function set as Python-Telegram-Bot MessageHandler for default text messages (this is the function called echo in the tutorial)
    • In case you reimplement your intents for Google/Alexa, you may skip implementing this function (since it would decompose into many different functions). We'll check manually 🙂.
  • Add a short README telling us how to run your bot. You don't need to commit any API tokens (we can get our own for testing), but let us know in the readme where to add the token.


  • Heroku is a free service that allows you to deploy your apps. See the tutorial. Deploying is optional.
  • Alternatively, you can use ngrok for testing purposes.
  • To import dialmonkey.* in your solution so that you can easily use your own modules, there are essentially two ways:
    • You can make an editable install of your Dialmonkey-npfl123 checkout directory -- within your virtualenv/conda environment, simply go to your checkout directory and run:
pip install -e .

That way, you can make imports from all the Dialmonkey packages from anywhere.

  • Alternatively, you can add the .. directory into hw9/'s sys.path. This is essentially what we're doing with the automatic tests. Use relative paths for the import, like this:
import os, sys
sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))

10. Grapheme-to-phoneme conversion

 Presented: 24 April, Deadline: 18 May

This time, your task will be to create a grapheme-to-phoneme conversion that works with the MBROLA concatenative speech synthesis system. By default, we'll assume you'll use Czech or English with this homework (only one of them, your choice :-)). If you want to try it out for a different language instead, we're open to that, but please talk to us first and check if MBROLA supports it.


  1. Install MBROLA. On Debian-based Linuxes, this should be as simple as sudo apt install mbrola. Building for any other Linux shouldn't be too hard either. On Windows, you can use WSL, or you can get native Windows binaries here.

  2. Install a MBROLA voice for your language. On Debian-derivatives (incl. WSL), you can go with sudo apt install mbrola-<voice> and your voices will into /usr/share/mbrola, otherwise you just need to download the voice somewhere.

    • For Czech, cz2 is a good voice, cz1 is lacking some rather common diphone combinations.

    • For English, you can go with en1.

    You can try out that it's working by running MBROLA through one of the test files included with the voice. There's always a “test” subdirectory with some “.pho” files.

    mbrola -e /path/to/cz2  path/to/test/some_file.pho output.wav
  3. Implement a simple normalization script. It should be able to expand numbers (just single digits) and abbreviations from a list.

    • Ignore the fact that you sometimes need context for the abbreviations.
    • Add the following abbreviations to your list to test it: Dr, Prof, kg, km, etc/atd.
  4. Add a grapheme-to-phoneme conversion to your script that produces a phoneme sequence like this:


    a    100
    h\   50
    o    70
    j    50
    _    200


    h    50
    @    70
    l    50
    @U   200
    _    200

    It's basically a two-column tab/space-separated file. The first column is a phoneme, the 2nd column denotes the duration in milliseconds.

    The available phonemes for each language are defined in the voices' README files (cs, en). MBROLA uses the SAMPA phonetic notation. The _ denotes a pause in any language.

    Use the following simple rules for phoneme duration:

    • Consonant – 50 ms
    • Short vowel (any vowel without “:” in Czech SAMPA, any 1-character vowel in English SAMPA): stressed – 100 ms, unstressed – 70 ms
    • Long vowel (vowels with “:” in Czech SAMPA, 2-character vowels in English SAMPA): stressed – 200 ms, unstressed – 150 ms

    If you inspect the MBROLA test files or the description here, you'll see that there's an optional third column for voice melody, saying which way F0 should develop during each phoneme. For our exercise, we'll ignore it. It'll give you a rather robotic, but understandable voice. What you should do, though is:

    • Add a 200 ms pause after each comma or dash.
    • Add a 500 ms pause after sentence-final punctuation (full stop, exclamation or question mark).
    • Do not add any pauses between words, since you also don't make them when speaking :-).

    Finally, the actual grapheme-to-phoneme rules are very different for both languages.

    • For Czech, you can do almost everything by starting from ortography and applying some relatively simple rules.

      • You should also add a dictionary for exceptions – include these 7 foreign words with their correct SAMPA pronunciations, to test that it works correctly: business, diesel, design, interview, Newton, pizza, revue
    • For English, you can't do without a dictionary. Use the CMU Pronouncing Dictionary, which you can find in data/hw10/cmudict.dict in your repo if you update from upstream.

      • Since the dictionary uses Arpabet and you want SAMPA for MBROLA, you'll need to create an Arpabet-to-SAMPA mapping to use it.
      • The dictionary has stress marks (“1”, “2”, “3” etc. after vowels, so you can treat vowels with “1” or “2” as stressed, the rest as unstressed).
      • Let the system spell out any word that it doesn't find in the dictionary (get the pronunciation of each letter).
  5. Take a random Wikipedia article (say, “article of the day”) in your target language, produce the g2p conversion for the first paragraph, then run MBROLA on it and save a WAV file.

Implementation notes

To make our tests easier, please include a function called g2p_czech(text) or g2p_english(text) depending on your target language, which takes plain text on the input and outputs a list of tuples (sound, duration). For instance, for Speech., we get: [('s', 50), ('p', 50), ('i:', 200), ('tS', 50), ('_', 500)].

Note that you can easily convert this to the target text format: '\n'.join([sound + '\t' + str(duration) for sound, duration in output]).

What to include in your merge request

Create a directory hw10/ and put into it:

  • hw10/ -- Your normalization and grapheme-to-phoneme Python script that will take plain text input and outputs a MBROLA instructions file.
    • Make it read standard input and write to standard output. Use if __name__ == '__main__' so we can import the file for testing your g2p_<czech|english> function.
  • hw10/test.txt -- The text of the paragraph on which you tried your conversion system.
  • hw10/test.pho -- Your script's output on the paragraph (the “.pho” file for MBROLA).
  • hw10/test.wav -- Your resulting MBROLA-produced WAV file.

11. Digits ASR

 Presented: 15 May, Deadline: 8 June (extended)

This time, your task will be to train and evaluate an ASR system on a small dataset. We will be using pretrained Wav2vec 2.0 model from Huggingface and finetuning it further. This is an end-to-end ASR model, so we won't have to fiddle with language models or pronunciation dictionary -- it goes straight from audio to characters.

This assignment is relatively compute-heavy. You can use Google Colab for the task, or any machine with a GPU. While it techncially could run on a CPU, in practice it doesn't, for a reason that Ondrej can't figure out 😳. If you don't want to work with Google Colab or a GPU, you can try out the 2021 assignment with Kaldi -- that's not an end-to-end, with separate acoustic and language models, and trains fine on a CPU, but it's really fiddly to set it up!


Basically, what we want you to do is take the pretrained Wav2Vec2 model, measure WER on Ondrej's fork of the Free Spoken Digits Dataset, then finetune it for a random data split and a per-speaker split, and compare and analyze the results.

You have two options -- you can either use the supplied Python code (best for working on a cluster), or you can use the IPython notebook (best for Google Colab).

Detailed Instructions

  1. Check and examine the given code (either Python or IPython notebok) + library of functions + Wav2vec2 online description, so you know how to complete the code.

  2. Set up Wav2vec2 required libraries (see requirements.txt) -- note that they'll take up a few gigs of space. Also download Ondrej's fork of the Free Spoken Digits Dataset. This is included in the IPython notebook as well.

    • Note that the data has two train-test splits: random and per speaker
    • Technical note: the audio has 8kHz sampling rate, but Wav2Vec2 uses 16kHz internally. There's a loading function in the library that deals with this.
  3. Implement:

    • a function for running the predictions (see the Wav2Vec2 online description for this)
    • a function for loading the data from a given directory (use the library function and assign transcripts, note that the target transcriptions need to be uppercased!)
    • a function for computing WER (see the Wav2Vec2 online description for this)
  4. Load the non-finetuned model (see Wav2Vec2 description), input processor, and data collator.

  5. Load the data for random split and compute WER with the non-finetuned model.

  6. Using a function from the library (which includes all needed settings), finetune the model on the random data split. Measure WER again.

  7. Reload the non-finetuned model from scratch (don't forget this, it's important!).

  8. Load the data for the per-speaker split. Compute non-finetuned model WER on this data.

  9. Finetune the model on this split.

  10. Measure WER + find most frequently misheard digits (print outputs vs. labels and compare manually).

Files to include in your merge request

  • Your completed code for the finetuning, prediction and WER measurement in either hw11/ or hw11/hw11.ipynb -- only one of these is needed, not both :-).
  • A Markdown file hw11/, which should contain:
    • Your WER for four settings:
      • Random split, non-finetuned model
      • Random split, finetuned model
      • Per-speaker split, non-finetuned model
      • Per-speaker split, finetuned model
    • A short report -- please try to explain:
      • why the WER results ended up the way they did,
      • which digits are most difficult to recognize for the per-speaker finetuned model and why.

12. Retrieval chatbot

 Presented: 22 May, Deadline: 15 June (extended)

This time, you will implement a basic information-retrieval-based chatbot. We'll just use TF-IDF for retrieval, with no smart reranking.


  1. We'll use the DailyDialog dataset. This is a dataset of basic day-to-day dialogues, so it's great for chatbots. It's already inclued in your repository, under data/hw12. Have a look at the data format, it's not very hard to parse.

  2. Implement an IR chatbot module (recommended approach, alternatives and extensions welcome):

    • Load the DailyDialog data into memory so that you know which turn follows which. Use data/hw12/dialogues_train.txt for this. Note that you shouldn't load the file at every turn -- load it once in the constructor. Please use relative paths (see HW8 notes).

    • From your data, create a Keys dataset, containing all turns in all dialogues except the last one. Then create a Values dataset, which always contain the immediately next turn for each dialogue.

      • Say there's just 1 dialogue with 5 turns (represented just by numbers here). Keys should contain [0, 1, 2, 3] and the corresponding Values are [1, 2, 3, 4].
    • Use TfidfVectorizer from Scikit-Learn as the main “engine”.

      • Create a vectorizer object and call fit_transform on the Keys set to train your matching TF-IDF matrix (store this matrix for later). Feel free to play around with this method's parameters, especially with the ngram_range -- setting it slightly higher than the default (1,1) might give you better results.
    • For any input sentence, what your chatbot should do is:

      • Call transform on your vectorizer object to obtain TF-IDF scores.

      • Get the cosine similarity of the sentence's TF-IDF to all the items in the Keys dataset.

      • Find the top 10 Keys' indexes using numpy.argpartition (see the example here). Now get the corresponding top 10 Values (at the same indexes). Choose one of them at random and use it as output.

        • Instead of choosing at random, you could do some smart reranking, but we'll skip that in this exercise.
  3. Integrate your chatbot into DialMonkey. Create a module inside dialmonkey.policy and add a corresponding YAML config file. You can call it ir_chatbot.

    • Note that you can create a config file where this module will be the only thing, essentially standing in for NLU, policy and NLG at the same time. You just access the dial.user input and directly call dial.set_system_response().
  4. Take the 1st sentence of the first 10 DailyDialog validation dialogues (data/hw12/dialogues_validation.txt) and see what your chatbot tells you.

Files to include in your merge request

  • Your chatbot module under dialmonkey/policy/ and your configuration file under conf/ir_chatbot.yaml.
  • The first 10 DailyDialog validation opening lines along with your chatbot's responses under hw12/samples.txt.

Further reading

More low-level stuff on TF-IDF:

Homework Submission Instructions

All homework assignments will be submitted using a Git repository on MFF GitLab.

We provide an easy recipe to set up your repository below:

Creating the repository

  1. Log into your MFF gitlab account. Your username and password should be the same as in the CAS, see this.
  2. You'll have a project repository created for you under the teaching/NPFL123/2023 group. The project name will be the same as your CAS username. If you don't see any repository, it might be the first time you've ever logged on to Gitlab. In that case, Ondřej first needs to run a script that creates the repository for you (please let him know on Slack). In any case, you can explore everything in the base repository. Your own repo will be derived from this one.
  3. Clone your repository.
  4. Change into the cloned directory and run
git remote show origin

You should see these two lines:

* remote origin
  Fetch URL:
  Push  URL:

  1. Add the base repository (with our code, for everyone) as your upstream:
git remote add upstream
  1. You're all set!

Submitting the homework assignment

  1. Make sure you're on your master branch
git checkout master
  1. Checkout a new branch -- make sure to name it hwXX (e.g. “hw4” or “hw11”) so our automatic checks can find it later!
git checkout -b hwXX
  1. Solve the assignment :)

  2. Add new files and commit your changes -- make sure to name your files as required, or you won't pass our automatic checks!

git add hwXX/
git commit -am "commit message"
  1. Push to your origin remote repository:
git push origin hwXX
  1. Create a Merge request in the web interface. Make sure you create the merge request into the master branch in your own forked repository (not into the upstream).

     Merge requests -> New merge request
Merge request
  1. Wait a bit till we check your solution, then enjoy your points :)!
  2. Once approved, merge your changes into your master branch – you might need them for further homeworks.

Updating from the base repository

You might need to update from the upstream base repository every once in a while (most probably before you start implementing each assignment). We'll let you know when we make changes to the base repo.

To upgrade from upstream, do the following:

  1. Make sure you're on your master branch
git checkout master
  1. Fetch the changes
git fetch upstream master
  1. Apply the diff
git merge upstream/master master

Exam Question Pool

The exam will have 10 questions, mostly from this pool. Each counts for 10 points. We reserve the right to make slight alterations or use variants of the same questions. Note that all of them are covered by the lectures, and they cover most of the lecture content. In general, none of them requires you to memorize formulas, but you should know the main ideas and principles. See the Grading tab for details on grading.


  • What's the difference between task-oriented and non-task-oriented systems?
  • Describe the difference between closed-domain, multi-domain, and open-doman systems.
  • Describe the difference between user-initiative, mixed-initiative, and system-initiative systems.

Data & Evaluation

  • What are the typical options for collecting dialogue data?
  • How does Wizard-of-Oz data collection work?
  • What is corpus annotation, what is inter-annotator agreement?
  • What is the difference between intrinsic and extrinsic evaluation?
  • What is the difference between subjective and objective evaluation?
  • What are the main extrinsic evaluation techniques for task-oriented dialogue systems?
  • What are some evaluation metrics for non-task-oriented systems (chatbots)?
  • What's the main metric for evaluating ASR systems?
  • What's the main metric for NLU (both slots and intents)?
  • Explain an NLG evaluation metric of your choice.
  • Why do you need to check for statistical significance (when evaluating an NLP experiment and comparing systems)?
  • Why do you need to evaluate on a separate test set?

Linguistics of Dialogue

  • What are turn taking cues/hints in a dialogue? Name a few examples.
  • Explain the main idea of the speech acts theory.
  • What is grounding in dialogue?
  • Give some examples of grounding signals in dialogue.
  • What is deixis? Give some examples of deictic expressions.
  • What is coreference and how is it used in dialogue?
  • What does Shannon entropy and conditional entropy measure? No need to give the formula, just the principle.
  • What is alignment/entrainment in dialogue?

Natural Language Understanding

  • What are some alternative semantic representations of utterances, in addition to dialogue acts?
  • Describe language understanding as classification and language understanding as sequence tagging.
  • How do you deal with conflicting slots or intents in classification-based NLU?
  • What is delexicalization and why is it helpful in NLU?
  • Describe one of the approaches to slot tagging as sequence tagging.
  • What is the IOB/BIO format for slot tagging?
  • What is the label bias problem?
  • How can an NLU system deal with noisy ASR output? Propose an example solution.

Neural NLU & Dialogue State Tracking

  • Describe a neural architecture for NLU.
  • What is the dialogue state and what does it contain?
  • What is an ontology in task-oriented dialogue systems?
  • Describe the task of a dialogue state tracker.
  • What's a partially observable Markov decision process?
  • Describe a viable architecture for a belief tracker.
  • What is the difference between dialogue state and belief state?
  • What's the difference between a static and a dynamic state tracker?

Dialogue Policies

  • What are the non-statistical approaches to dialogue management/action selection?
  • Why is reinforcement learning preferred over supervised learning for training dialogue managers?
  • Describe the main idea of reinforcement learning (agent, environment, states, rewards).
  • What are deterministic and stochastic policies in dialogue management?
  • What's a value function in a reinforcement learning scenario?
  • What's the difference between actor and critic methods in reinforcement learning?
  • What's the difference between model-based and model-free approaches in RL?
  • What are the main optimization approaches in reinforcement learning?
  • Why do you typically need a user simulator to train a reinforcement learning dialogue policy?

Neural Policies & Natural Language Generation

  • How do you involve neural networks in reinforcement learning (describe a Q network or a policy network)?
  • What are the main steps of a traditional NLG pipeline – describe at least 2.
  • Describe one approach to NLG of your choice.
  • Describe how template-based NLG works.
  • What are some problems you need to deal with in template-based NLG?
  • Describe a possible neural networks based NLG architecture.

Voice assistants & Question Answering

  • What is a smart speaker made of and how does it work?
  • Briefly describe a viable approach to question answering.
  • What is document retrieval and how is it used in question answering?
  • What is a knowledge graph?

Dialogue Tooling

  • What is a dialogue flow?
  • What are intents and entities/slots?
  • How can you improve a chatbot in production?

Text-to-speech Synthesis

  • How do humans produce sounds of speech?
  • What's the difference between a vowel and a consonant?
  • What is F0 and what are formants?
  • What is a spectrogram?
  • What are main distinguishing characteristics of consonants?
  • What is a phoneme?
  • What are the main distinguishing characteristics of different vowel phonemes (both how they're produced and perceived)?
  • What are the main approaches to grapheme-to-phoneme conversion in TTS?
  • Describe the main idea of concatenative speech synthesis.
  • Describe the main ideas of statistical parametric speech synthesis.
  • How can you use neural networks in speech synthesis?

Automatic Speech Recognition

  • What is a speech activity detector?
  • Describe the main components of an ASR pipeline system.
  • How do input features for an ASR model look like?
  • What is the function of the acoustic model in a pipeline ASR system?
  • What's the function of a decoder/language model in a pipeline ASR system?
  • Describe the architecture of an end-to-end neural ASR system.


  • What are the three main approaches to building chatbots?
  • How does the Turing test work? Does it have any weaknesses?
  • What are some techniques rule-based chatbots use to convince their users that they're human-like?
  • Describe how a retrieval-based chatbot works.
  • How can you use neural networks for chatbots? Does that have any problems?
  • Describe a possible architecture of an ensemble chatbot.

Course Grading

To pass this course, you will need to:

  1. Take an exam (a written test covering important lecture content).
  2. Do lab homeworks (various dialogue system implementation tasks).

Exam test

  • There will be a written exam test at the end of the semester.
  • There will be 10 questions, we expect 2-3 sentences as an answer, with a maximum of 10 points per question.
  • To pass the course, you need to get at least 50% of the total points from the test.
  • We plan to publish a list of possible questions beforehand.

In case the pandemic does not get better by the exam period, there will be a remote alternative for the exam (an essay with a discussion).

Homework assignments

  • There will be 12 homework assignments, introduced every week, starting on the 2nd week of the semester.
  • You will submit the homework assignments into a private Gitlab repository (where we will be given access).
  • For each assignment, you will get a maximum of 10 points.
  • All assignments will have a fixed deadline.
  • If you submit the assignment after the deadline, you will get:
    • up to 50% of the maximum points if it is less than 2 weeks after the deadline;
    • 0 points if it is more than 2 weeks after the deadline.
  • Once we check the submitted assignments, you will see the points you got and the comments from us as comments on your merge requests on Gitlab.
  • To be allowed to take the exam (which is required to pass the course), you need to get at least 50% of the total points from the assignments.


The final grade for the course will be a combination of your exam score and your homework assignment score, weighted 3:1 (i.e. the exam accounts for 75% of the grade, the assignments for 25%).


  • Grade 1: >=87% of the weighted combination
  • Grade 2: >=74% of the weighted combination
  • Grade 3: >=60% of the weighted combination
  • An overall score of less than 60% means you did not pass.

In any case, you need >50% of points from the test and >50% of points from the homeworks to pass. If you get less than 50% from either, even if you get more than 60% overall, you will not pass.

No cheating

  • Cheating is strictly prohibited and any student found cheating will be punished. The punishment can involve failing the whole course, or, in grave cases, being expelled from the faculty.
  • Discussing homework assignments with your classmates is OK. Sharing code is not OK (unless explicitly allowed); by default, you must complete the assignments yourself.
  • All students involved in cheating will be punished. E.g. if you share your assignment with a friend, both you and your friend will be punished.

Recommended Reading

You should pass the course just by following the lectures, but here are some hints on further reading. There's nothing ideal on the topic as this is a very active research area, but some of these should give you a broader overview.

Basic (good but very brief, available online):

More detailed (very good, available as e-book from our library):

Further reading:

  • Janarthanam: Hands-On Chatbots and Conversational UI Development. Packt 2017.
    • practical guide on developing dialogue systems for current platforms, virtually no theory
  • Gao et al.: Neural Approaches to Conversational AI. arXiv:1809.08267
    • an advanced, good overview of the latest neural approaches in dialogue systems
  • McTear et al.: The Conversational Interface: Talking to Smart Devices. Springer 2016.
    • practical, for current platforms, more advanced and more theory than Janarthanam
  • Jokinen & McTear: Spoken dialogue systems. Morgan & Claypool 2010.
    • good but slightly outdated, some systems very specific to particular research projects
  • Rieser & Lemon: Reinforcement learning for adaptive dialogue systems. Springer 2011.
    • advanced, slightly outdated, project-specific
  • Lemon & Pietquin: Data-Driven Methods for Adaptive Spoken Dialogue Systems. Springer 2012.
    • ditto
  • Skantze: Error Handling in Spoken Dialogue Systems. PhD Thesis 2007, Chap. 2.
    • good introduction into dialogue systems in general, albeit slightly dated
  • McTear: Spoken Dialogue Technology. Springer 2004.
    • good but dated
  • Psutka et al.: Mluvíme s počítačem česky. Academia 2006.
    • virtually the only book in Czech, good for ASR but dated, not a lot about other parts of dialogue systems