ConvAI3: Clarifying Questions for Open-Domain Dialogue Systems (ClariQ) by DeepPavlov_Home & Family

ConvAI3: Clarifying Questions for Open-Domain Dialogue Systems (ClariQ) SCAI workshop @ EMNLP 2020 View On GitHub NEW! Challenge winners announced!ClariQ OverviewClariQ (pronounced as Claire-ee-que) challenge is organized as part of the Conversational AI challenge series (ConvAI3) at Search-oriented Conversational AI (SCAI) EMNLP workshop in 2020.The main aim of the conversational systems is to return an appropriate answer in response to the user requests. However, some user requests might be ambiguous. In IR settings such a situation is handled mainly through the diversification of a search result page.It is however much more challenging in dialogue settings. Hence, we aim to study the following situation for dialogue settings: a user is asking an ambiguous question (a question to which one can return > 1 possible answers); the system must identify that the question is ambiguous, and, instead of trying to answer it directly, ask a good clarifying question.The main research questions we aim to answer as part of the challenge are the following: RQ1: When to ask clarifying questions during dialogues? RQ2: How to generate the clarifying questions?The detailed description of the challenge can be found in the following document.How to participate? In order to get to the leaderboard please register your team here. The datasets, baselines and evaluation scripts for the Stage 1 are available in the following repository https://github.com/aliannejadi/ClariQ Email us if you have any questions News June 10, 2020: Microsoft Research is sponsoring data collection for the Stage 1 July 7, 2020: Announcing the Stage 1 of ClariQ challenge July 8, 2020: Announcing the performance of the baselines July 20, 2020: Send us your solution for Stage 1 every Friday via email July 25, 2020: Ask us question July 30, 2020: Please join our kick-off webinar on August 4, 8 am PST (convert to your time zone) using the following meeting link, where we give a brief overview of the challenge; show how to run baselines; answer all your questions. July 30, 2020: Don’t want to miss any updates please register here August 4, 2020: Google is sponsoring the competion prize (see the details below) August 4, 2020: We held the kick-off webinar. Missed it? You can watch the playback here. August 10, 2020: New baselines for the question_relevance have been added. The code is also available on Google Colab. September 08, 2020: Due to multiple requests, we have extended the deadline for participation in Stage 1 for one week. September 22, 2020: Amazon Science is sponsoring the Stage 2 of the challenge September 23, 2020: We are happy to announce four teams which were selected for the Stage 2: Karl NTES_ALONG TAL ML Soda September 23, 2020: Stage 2 files and instructions are added to the ClariQ GitHub repo. November 19, 2020: Stage 2 results are out. November 19, 2020: Challenge winners are announced.AwardsGoogle is sponsoring the competition award with the GCP Credits, which we are planning to award as follows: 1st place: $2’500 $3’000 credits 2nd place: $1’500 $2’000 credits 3rd place: $1’000 creditsIMPORTANT! Please note that GCP Credits can only be granted where available and are subject to the Terms and Conditions and product availability. If the competition winner is from a region where the program is not launched, we will, unfortunately, not be able to issue the prize.Challenge DesignThe challenge will be run in two stages:Stage 1: intial dataset (sponsored by Microsoft Research)In Stage 1, we provide to the participants the datasets that include: User Request: an initial user request in the conversational form, e.g., “What is Fickle Creek Farm?”, with a label reflects if clarification is needed to be ranged from 1 to 4; Clarification questions: a set of possible clarifying questions, e.g., “Do you want to know the location of fickle creek farm?”; User Answers: each question is supplied with a user answer, e.g., “No, I want to find out where can i purchase fickle creek farm products.”To answer RQ1: Given a user request, return a score [1 −4] indicating the necessity of asking clarifying questions.To answer RQ2: Given a user request which needs clarification, returns the most suitable clarifying question. Here participants are able to choose: (1) either select the clarifying question from the provided question bank (all clarifying questions we collected), aiming to maximize the precision, (2) orchoose not to ask any question (by choosing Q0001 from the question bank.)The dataset is stored in the following repository https://github.com/aliannejadi/ClariQ, together with evaluation scripts and baseline.Stage 2: human-in-the-loop (sponsored by Amazon Science)The TOP-4 systems from Stage 1 are exposed to real users. Their responses—answers and clarifying questions—are rated by the users.At that stage, the participating systems are put in front of human users. The systems are rated on their overall performance.At each dialog step, a system should give either a factual answer to the user’s query or ask for clarification.Therefore, the participants would need to: ensure their system can answer simple user questions make their own decisions on when clarification might be appropriate provide clarification question whenever appropriate interpret user’s reply to the clarifying questionThe participants would need to strike a balance between asking too many questionsand providing irrelevant answers.Note that the setup of this stage is quite different from Stage 1. Participating systems would likely need to operate as a generative model, rather than a retrieval model. One option would be to cast the problem as generative from the beginning and solve the retrieval part of Stage 1, e.g., by ranking the offered candidates by their likelihood.Alternatively, one may solve Stage 2 by retrieving a list of candidate answers (e.g., by invoking Wikipedia API or the Chat Noir API that we describe above) and ranking them as in Stage 1.For further information about Stage 2 design and auxiliary files, please visit ClariQ repo.Timeline Stage 1 will take place from July 7, 2020 – September 9, 2020 September 16, 2020. Up until September 9, 2020 September 16, 2020 participants will be able to submit their models (source code) and solutions to be evaluated on the test set using automated metrics (which we will run on our servers). The current leaderboards will be visible to everyone. Stage 2 will start on September 10, 2020 September 17 2020. On September 10, 2020 September 17 2020 the run submission for Stage 1 will be locked, and the best teams will be invited to submit their codes for the second stage to be evaluated over the next month using crowd workers.Winners will be announced at SCAI@EMNLP2020 which will take place in November 19-20 (exact details TBD).EvaluationParticipants’ models will then be compared in two ways after two stages: automated evaluation metrics on a new test set hidden from the competitors; evaluation with crowd workers through MTurk.The winning will be chosen based on these scores.MetricsThere are three types of metrics we will evaluate: Automated metrics As system automatic evaluation metrics we use MRR, P@[1,3,5,10,20], nDCG@[1,3,5,20].These metrics are computed as follows: a selected clarifying question, together with its corresponding answer are added to the original request.The updated query is then used to retrieve (or re-rank) documents from the collection. The quality of a question is then evaluated by taking into account how much the question and its answer affect the performance of document retrieval. Models are also evaluated in how well they are able to rank relevant questions higher than other questions in the question bank. For this task, that we call ‘question relevance’, the models are evaluated in terms of Recall@[10,20,30]. Since the precision of models is evaluated in the document relevance task, here we focus only on recall. Crowd workers given the entrant’s model code, we will run live experiments where Turkers chat to their model given instructions identical to the creation of the original dataset, but with new profiles, and then score its performance. Turkers will score the models between 1-5. Rules Participants must provide their source code so that the hidden test set evaluation and live experiments can be computed without the team’s influence, and so that the competition has further impact as those models can be released for future research to build off them. Code can be in any language but a thin python wrapper must be provided in order to work with our evaluation and live experiment code. We will require that the winning systems also release their training code so that their work is reproducible (although we also encourage that for all systems). Participants are free to augment training with other datasets as long as they are publicly released (and hence, reproducible). Hence, all entrants are expected to work on publicly available data or release the data they use to train.SubmissionStage 1Please send two files per run as to [email protected], indicating your team’s name, as well as your run ID. Each team is allowed to send a maximum of one run per week.You’ll also need to share your GitHub repository with us. The run files should be formatted as described below.Run file formatEach run consists of two separate files: Ranked list of questions for each topic; Predicted clarification_need label for each topic.Below we explain how each file should be formatted.Question rankingThis file is supposed to contain a ranked list of questions per topic. The number of questions per topic could be any number, but we evaluate only the top 30 questions. We follow the traditional TREC run format. Each line of the file should be formatted as follows: 0 Each line represents a relevance prediction. is the relevance score that a model predicts for a given and . is a string indicating the ID of the submitted run. denotes the ranking of the for . Practically, the ranking is computed by sorting the questions for each topic by their relevance scores.Here are some example lines:170 0 Q00380 1 6.53252 sample_run170 0 Q02669 2 6.42323 sample_run170 0 Q03333 3 6.34980 sample_run171 0 Q03775 1 4.32344 sample_run171 0 Q00934 2 3.98838 sample_run171 0 Q01138 3 2.34534 sample_runThis run file will be used to evaluate both question relevance and document relevance. Sample runs can found in ./sample_runs/ directory.Also, sample Google Colab Notebookes are available. Please check ClariQ repo for more information.Clarification needThis file is supposed to contain the predicted clarification_need labels. Therefore, the file format is simply the topic_id and the predicted label. Sample lines can be found below:171 1170 3182 4More information and example run files can be found at https://github.com/aliannejadi/ClariQ.Stage 2To submit an entry, create a private repo with your model that works with our evaluation code, and share it with the following github accounts: aliannejadi, varepsilon, julianakiseleva.See https://github.com/aliannejadi/ClariQ for example baseline submissions.You are free to use any system (e.g. PyTorch, Tensorflow, C++,..) as long as you can wrap your model for the evaluation. The top level README should tell us your team name, model name, and where the eval_ppl.py, eval_hits.py etc. files are so we can run them. Those should give the numbers on the validation set. Please also include those numbers in the README so we can check we get the same. We will then run the automatic evaluations on the hidden test set and update the leaderboard. You can submit a maximum of once per week.We will use the same submitted code for the top performing models for computing human evaluations when the submission system is locked on September 9, 2020 September 17, 2020. The deadline for submitting systems for Stage 2 is November 1, 2020.The submitted systems must produce the output in a reasonable time window (1-2 hours) on our system with the following configuration: CPU: Core i7-7700 RAM: 32 Gb RAM GPU: 1080 TiNote: All submitted systems must be accompanied with a ready-to-use Docker container.Automatic Evaluation Leaderboard (hidden test set)Document RelevanceDev ًRank Creator Model Name MRR P@1 nDCG@3 nDCG@5 - ClariQ Oracle BestQuestion 0.4882 0.4187 0.3337 0.3064 1 Karl Roberta 0.3640 0.2813 0.2002 0.1954 2 NTES_ALONG Reranker-v4 0.3761 0.3000 0.2113 0.1955 3 Soda BERT+BM25-v2 0.3180 0.2437 0.1625 0.1550 4 NTES_ALONG Reranker-v2 0.3573 0.2625 0.2112 0.1982 5 ClariQ BM25 0.3096 0.2313 0.1608 0.1530 6 Soda BERT+BM25 0.3096 0.2313 0.1608 0.1530 7 ClariQ NoQuestion 0.3000 0.2063 0.1475 0.1530 8 Pinta BERT-v3 0.3596 0.2750 0.1879 0.1882 9 NTES_ALONG BM25+Roberta 0.3606 0.2813 0.1942 0.1891 10 CogIR BERTlets-fusion-topics-passages 0.3103 0.2125 0.1747 0.1701 11 Pinta BERT 0.3297 0.2250 0.1792 0.1701 12 CogIR BERTlets-fusion-topics-div-passages-v2 0.3315 0.2500 0.1763 0.1660 13 CogIR BERTlets-fusion-topics-div-passages 0.3236 0.2250 0.1739 0.1653 14 Pinta BERT-v2 0.3158 0.2313 0.1669 0.1600 15 Karl Roberta-v2 0.3811 0.2938 0.2193 0.2093 16 NTES_ALONG Reranker-v3 0.3520 0.2687 0.2033 0.1925 17 NTES_ALONG Recall+Rerank 0.3627 0.2750 0.2047 0.1935 18 Soda BERT-based-v2 0.3306 0.2437 0.1699 0.1702 19 Soda BERT-based 0.3497 0.2625 0.1849 0.1762 20 Algis USE-QA 0.3517 0.2563 0.1943 0.1815 21 Pinta Triplet 0.3573 0.2688 0.1988 0.1920 22 TAL ML Roberta+++ 0.3619 0.2750 0.2060 0.1979 23 NTES_ALONG Recall+Rescore 0.3722 0.2813 0.2185 0.2047 24 NTES_ALONG BM25_plus+Roberta 0.3587 0.2813 0.1952 0.1869 25 Algis BART-based 0.3628 0.2687 0.2003 0.1914 26 TAL ML Roberta++ 0.3583 0.2687 0.1977 0.1931 27 ClariQ BERT-ranker 0.3453 0.2563 0.1824 0.1744 28 ClariQ BERT-reranker 0.3453 0.2563 0.1824 0.1744 - ClariQ Oracle WorstQuestion 0.0841 0.0125 0.0252 0.0313 Test ًRank Creator Model Name MRR P@1 nDCG@3 nDCG@5 - ClariQ Oracle BestQuestion 0.4881 0.4275 0.2107 0.1759 1 Karl Roberta 0.3190 0.2342 0.1265 0.1130 2 NTES_ALONG Reranker-v4 0.3140 0.2379 0.1229 0.1097 3 Soda BERT+BM25-v2 0.3216 0.2453 0.1196 0.1097 4 NTES_ALONG Reranker-v2 0.3034 0.2119 0.1171 0.1033 5 ClariQ BM25 0.3134 0.2193 0.1151 0.1061 6 Soda BERT+BM25 0.3134 0.2193 0.1151 0.1061 7 ClariQ NoQuestion 0.3223 0.2268 0.1134 0.1059 8 Pinta BERT-v3 0.3044 0.2119 0.1131 0.1021 9 NTES_ALONG BM25+Roberta 0.3045 0.2156 0.1108 0.1025 10 CogIR BERTlets-fusion-topics-passages 0.3025 0.2193 0.1078 0.0983 11 Pinta BERT 0.2934 0.2045 0.1078 0.0969 12 CogIR BERTlets-fusion-topics-div-passages-v2 0.2885 0.1859 0.1072 0.1010 13 CogIR BERTlets-fusion-topics-div-passages 0.2908 0.1970 0.1055 0.0990 14 Pinta BERT-v2 0.2815 0.1933 0.1043 0.0934 15 Karl Roberta-v2 0.2890 0.1933 0.1035 0.0941 16 NTES_ALONG Reranker-v3 0.3006 0.2230 0.1031 0.0970 17 NTES_ALONG Recall+Rerank 0.2948 0.1933 0.1029 0.0919 18 Soda BERT-based-v2 0.2803 0.1896 0.1021 0.0981 19 Soda BERT-based 0.2600 0.1784 0.0983 0.0915 20 Algis USE-QA 0.2782 0.1822 0.0978 0.1003 21 Pinta Triplet 0.2672 0.1747 0.0968 0.0906 22 TAL ML Roberta+++ 0.2835 0.2007 0.0965 0.0915 23 NTES_ALONG Recall+Rescore 0.2799 0.1970 0.0955 0.0856 24 NTES_ALONG BM25_plus+Roberta 0.2720 0.1822 0.0930 0.0870 25 Algis BART-based 0.2622 0.1710 0.0923 0.0848 26 TAL ML Roberta++ 0.2602 0.1747 0.0922 0.0833 27 ClariQ BERT-ranker 0.2562 0.1784 0.0896 0.0821 28 ClariQ BERT-reranker 0.2553 0.1784 0.0892 0.0818 - ClariQ Oracle WorstQuestion 0.0541 0.0000 0.0097 0.0154 Question RelevanceDev ًRank Creator Model Name Recall@5 Recall@10 Recall@20 Recall@30 1 NTES_ALONG Reranker-v4 0.3604 0.6749 0.8478 0.8761 2 NTES_ALONG Reranker-v3 0.3648 0.6753 0.8510 0.8744 3 NTES_ALONG Reranker-v2 0.3648 0.6738 0.8417 0.8633 4 CogIR BERTlets-fusion-topics-div-passages-v2 0.3542 0.6424 0.7653 0.7997 5 TAL ML Roberta++ 0.3649 0.6694 0.8265 0.8587 6 Karl Roberta-v2 0.3611 0.6539 0.7993 0.8384 7 CogIR BERTlets-fusion-topics-passages 0.3555 0.6429 0.7640 0.7854 8 Karl Roberta 0.3618 0.6631 0.8128 0.8434 9 Soda BERT+Bm25 0.3454 0.6166 0.7354 0.7621 10 Soda BERT+Bm25-v2 0.3398 0.6166 0.7525 0.7792 11 Soda BERT-based 0.3523 0.6247 0.7354 0.7636 12 NTES_ALONG Recall+Rerank 0.3674 0.6678 0.7869 0.8085 13 Soda BERT-based-v2 0.3544 0.6287 0.7544 0.8177 14 Pinta BERT 0.3492 0.6196 0.7337 0.7632 15 CogIR BERTlets-fusion-topics-div-passages 0.3528 0.6393 0.7506 0.7890 16 NTES_ALONG BM25_plus+Roberta 0.3637 0.6409 0.7484 0.7793 17 Pinta BERT-v3 0.3583 0.6358 0.7616 0.7852 18 NTES_ALONG Recall+Rescore 0.3648 0.6553 0.8230 0.8367 19 ClariQ BERT-ranker 0.3494 0.6134 0.7248 0.7542 20 Pinta BERT-v2 0.3528 0.6393 0.7506 0.7890 21 Algis BART-based 0.3333 0.5910 0.6689 0.6926 22 Algis USE-QA 0.3469 0.6112 0.7052 0.7228 23 NTES_ALONG BM25+Roberta 0.3629 0.6389 0.7285 0.7657 24 TAL ML Roberta+++ 0.3508 0.6215 0.7672 0.7978 25 ClariQ BERT-reranker 0.3475 0.6122 0.6913 0.6913 26 ClariQ BM25 0.3245 0.5638 0.6675 0.6913 27 Pinta Triplet 0.3471 0.5871 0.6653 0.6846 Test ًRank Creator Model Name Recall@5 Recall@10 Recall@20 Recall@30 1 NTES_ALONG Reranker-v4 0.3404 0.6329 0.8335 0.8744 2 NTES_ALONG Reranker-v3 0.3414 0.6351 0.8316 0.8721 3 NTES_ALONG Reranker-v2 0.3382 0.6242 0.8177 0.8685 4 CogIR BERTlets-fusion-topics-div-passages-v2 0.3384 0.6314 0.8073 0.8573 5 TAL ML Roberta++ 0.3395 0.6251 0.8176 0.8568 6 Karl Roberta-v2 0.3355 0.6237 0.7990 0.8492 7 CogIR BERTlets-fusion-topics-passages 0.3314 0.6149 0.8074 0.8448 8 Karl Roberta 0.3406 0.6255 0.8006 0.8436 9 Soda BERT+BM25 0.3272 0.6061 0.8013 0.8433 10 Soda BERT+BM25-v2 0.3013 0.5866 0.8006 0.8433 11 Soda BERT-based 0.3338 0.6099 0.8023 0.8432 12 NTES_ALONG Recall+Rerank 0.3435 0.6296 0.7959 0.8424 13 Soda BERT-based-v2 0.3067 0.5893 0.7991 0.8415 14 Pinta BERT 0.3438 0.6228 0.7987 0.8409 15 CogIR BERTlets-fusion-topics-div-passages 0.3333 0.6225 0.8039 0.8392 16 NTES_ALONG BM25_plus+Roberta 0.3361 0.6219 0.7960 0.8360 17 Pinta BERT-v3 0.3291 0.6066 0.7902 0.8345 18 NTES_ALONG Recall+Rescore 0.3432 0.6229 0.7857 0.8211 19 ClariQ BERT-ranker 0.3440 0.6242 0.7849 0.8190 20 Pinta BERT-v2 0.3271 0.6012 0.7884 0.8188 21 Algis BART-based 0.3408 0.6156 0.7721 0.8081 22 Algis USE-QA 0.3454 0.6071 0.7688 0.8013 23 NTES_ALONG BM25+Roberta 0.3329 0.6027 0.7650 0.8004 24 TAL ML Roberta+++ 0.3212 0.5786 0.7204 0.7739 25 ClariQ BERT-reranker 0.3444 0.6062 0.7585 0.7682 26 ClariQ BM25 0.3170 0.5705 0.7292 0.7682 27 Pinta Triplet 0.3330 0.5809 0.7289 0.7589 Clarification Need Prediction ًRank Creator Model Name Dev Test P R F1 P R F1 1 TAL ML Roberta+++0.60390.56000.55510.59810.6557 0.6070 2 CactusJam Roberta+Stats 0.62000.58000.57170.59630.5902 0.5416 3 TAL ML Roberta++ 0.58070.54000.53750.52900.5574 0.5253 4 Algis Roberta+CatBoost 0.1402 0.2800 0.1854 0.5171 0.5246 0.5138 5 NTES_ALONG cneed_add_prior_v2 0.62000.60000.59840.50070.50820.5018 6 NTES_ALONG cneed_merge 0.5830 0.5200 0.5192 0.4847 0.50820.4960 7 NTES_ALONG cneed_dist 0.5452 0.5200 0.5177 0.4852 0.49180.4868 8 Karl Roberta-v2 0.46090.46000.43560.44650.5410 0.4871 9 NTES_ALONG Roberta+prior 0.4554 0.4600 0.4567 0.4926 0.4754 0.4799 10 Algis BartBoost0.70080.70000.69760.48130.47540.4756 11 Soda BERT-based-v2 0.5218 0.4800 0.4253 0.3931 0.4918 0.4350 12 Soda BERT-based 0.5224 0.5400 0.5180 0.3901 0.4754 0.4282 13 Soda BERT+BM25 0.5386 0.5600 0.5360 0.3930 0.47540.4273 14 Soda BERT+BM25-v2 0.5992 0.5800 0.5793 0.4304 0.42620.4207 15 Pinta BERT-v30.40830.50000.42480.37890.45900.4147 16 Pinta Triplet 0.4161 0.4800 0.4178 0.3665 0.45900.4074 17 Pinta BERT0.52040.50000.50140.39290.4098 0.4004 18 NTES_ALONG Roberta 0.3967 0.5200 0.4305 0.3546 0.4754 0.3995 19 Pinta XGB 0.50270.48000.48410.37760.3770 0.3771 Stage 2: Multi-turn Conversations EvalationDocument RelevanceTest ًRank Creator Model Name MRR P@1 nDCG@3 nDCG@5 1 NTES_ALONG ClariQ_Select_System 0.1798 0.1161 0.0553 0.0536 2 TAL ML MCAN 0.1669 0.1067 0.0522 0.0494 Challenge Winners First place: NTES_ALONG: Wenjie Ou, Yue Lin Second place: TAL ML: Hang Li, Yu Kang, Tianqiao Liu, Guowei Xu, Wenbiao Ding, Zitao LiuOrganizing team Mohammad Aliannejadi, University of Amsterdam, Amsterdam Julia Kiseleva, Microsoft Research, Seattle Mikhail Burtsev, MIPT, Moscow Aleksandr Chuklin, Google AI, Zürich Jeff Dalton, University of Glasgow, GlasgowPrevious ConvAI competitions ConvAI @ NIPS 2017 (The Conversational Intelligence Challenge) ConvAI2 @ NeurIPS 2018 (The Persona-Chat Challenge).SponsorsSponsorship of data collection:The Stage 1 of the challenge is sponsored by:The Stage 2 of the challenge is sponsored by:Participants prize is sponsored by: Project maintained by DeepPavlov Hosted on GitHub Pages — Theme by mattgraham