Deep Learning Methods �for Query Auto Completion� �
Manish Gupta
Meghana Joshi
Puneet Agrawal
{gmanish, mejoshi, punagr}@microsoft.com
Agenda
Manish Gupta (gmanish@microsoft.com)
2
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
3
DL for QAC, ECIR 23
AutoSuggest Examples
Manish Gupta (gmanish@microsoft.com)
4
DL for QAC, ECIR 23
Prefix
Block
Conversation
Conceptual difficulty of the AS problem
Manish Gupta (gmanish@microsoft.com)
5
DL for QAC, ECIR 23
apple
Intention Gap
Important Components in a QAC system
Manish Gupta (gmanish@microsoft.com)
6
DL for QAC, ECIR 23
Ranking suggestions: Most Popular Completion (MPC)
Whiting, Stewart, Andrew James McMinn, and Joemon M. Jose. "Exploring Real-Time Temporal Query Auto-Completion." In DIR, pp. 12-15. 2013.
Manish Gupta (gmanish@microsoft.com)
7
DL for QAC, ECIR 23
Ranking suggestions: Time sensitive suggestions
Shokouhi, Milad, and Kira Radinsky. "Time-sensitive query auto-completion." In SIGIR, pp. 601-610. 2012.
Wang, Yingfei, Hua Ouyang, Hongbo Deng, and Yi Chang. "Learning online trends for interactive query auto-completion." TKDE 29, no. 11 (2017): 2442-2454.
Manish Gupta (gmanish@microsoft.com)
8
DL for QAC, ECIR 23
Ranking suggestions: Location sensitive suggestions
Backstrom, Lars, Jon Kleinberg, Ravi Kumar, and Jasmine Novak. "Spatial variation in search engine queries." In WWW, pp. 357-366. 2008.
Welch, Michael J., and Junghoo Cho. "Automatically identifying localizable queries." In SIGIR, pp. 507-514. 2008.
Manish Gupta (gmanish@microsoft.com)
9
DL for QAC, ECIR 23
Ranking Suggestions: Personalization
Manish Gupta (gmanish@microsoft.com)
10
DL for QAC, ECIR 23
Using short-term/long-term user history, location, other signals
the great gatsby fitzgerald
the great gatsby book
the great influenza
the great alone book
the great gatsby film 2013
Search/Click History:
tender is the night
f. scott fitzgerald
this side of paradise
the silent patient
the great gatsby book
the great gatsby film 2013
the great gatsby trailer
the great gatsby film 1974
the greatest showman
the great wall film 2016
Search/Click History:
the wolf of wall street
the revenant
inception
gangs of new york
pain & gain
the great wall of china
the great gatsby film 2013
the great gatsby trailer
the great gatsby film 1974
the greatest showman
Search/Click History:
the wolf of wall street
the revenant
inception
gangs of new york
pain & gain
beijing travel advisory
forbidden city
Session history
Long-term
Ghosting, Session co-occurrences
Ramachandran, Lakshmi, and Uma Murthy. "Ghosting: contextualized query auto-completion on Amazon search." In SIGIR, pp. 1377-1378. 2019.
Bar-Yossef, Ziv, and Naama Kraus. "Context-sensitive query auto-completion." In WWW, pp. 107-116. 2011.
Manish Gupta (gmanish@microsoft.com)
11
DL for QAC, ECIR 23
Online spell correction, Defect handling
Small portions of the prefix can be corrected at trie exploration time paying a penalty cost. E.g. “cbo” 🡪 “ceboo”
Duan, Huizhong, and Bo-June Hsu. "Online spelling correction for query completion." In Proceedings of the 20th international conference on World wide web, pp. 117-126. 2011.
Manish Gupta (gmanish@microsoft.com)
12
DL for QAC, ECIR 23
Non-prefix matches, Generating suggestions
Gog, Simon, Giulio Ermanno Pibiri, and Rossano Venturini. "Efficient and effective query auto-completion." In SIGIR, pp. 2271-2280. 2020.
Manish Gupta (gmanish@microsoft.com)
13
DL for QAC, ECIR 23
Mobile QAC, Enterprise QAC
Zhang, Aston, Amit Goyal, Ricardo Baeza-Yates, Yi Chang, Jiawei Han, Carl A. Gunter, and Hongbo Deng. "Towards mobile query auto-completion: An efficient mobile application-aware approach." In WWW, pp. 579-590. 2016.
Hawking, David, and Kathy Griffiths. "An enterprise search paradigm based on extended query auto-completion: do we still need search and navigation?." In Proceedings of the 18th Australasian Document Computing Symposium, pp. 18-25. 2013.
Manish Gupta (gmanish@microsoft.com)
14
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
15
DL for QAC, ECIR 23
Ranking suggestions: Most Popular Completion (MPC)
Whiting, Stewart, Andrew James McMinn, and Joemon M. Jose. "Exploring Real-Time Temporal Query Auto-Completion." In DIR, pp. 12-15. 2013.
Manish Gupta (gmanish@microsoft.com)
16
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
17
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
18
DL for QAC, ECIR 23
Prefix and Suggestion Features
Sordoni, Alessandro, Yoshua Bengio, Hossein Vahabi, Christina Lioma, Jakob Grue Simonsen, and Jian-Yun Nie. "A hierarchical recurrent encoder-decoder for generative context-aware query suggestion." In CIKM, pp. 553-562. 2015.
Cai, Fei, and Maarten de Rijke. "Learning from homologous queries and semantically related terms for query auto completion." Information Processing & Management 52, no. 4 (2016): 628-643.
Manish Gupta (gmanish@microsoft.com)
19
DL for QAC, ECIR 23
Pairwise and Contextual Features
Sordoni, Alessandro, Yoshua Bengio, Hossein Vahabi, Christina Lioma, Jakob Grue Simonsen, and Jian-Yun Nie. "A hierarchical recurrent encoder-decoder for generative context-aware query suggestion." In CIKM, pp. 553-562. 2015.
Cai, Fei, and Maarten de Rijke. "Learning from homologous queries and semantically related terms for query auto completion." Information Processing & Management 52, no. 4 (2016): 628-643.
Manish Gupta (gmanish@microsoft.com)
20
DL for QAC, ECIR 23
Reformulation Features
Manish Gupta (gmanish@microsoft.com)
21
DL for QAC, ECIR 23
User Features
Li, Yanen, Anlei Dong, Hongning Wang, Hongbo Deng, Yi Chang, and ChengXiang Zhai. "A two-dimensional click model for query auto-completion." In SIGIR, pp. 455-464. 2014.
Di Santo, Giovanni, Richard McCreadie, Craig Macdonald, and Iadh Ounis. "Comparing approaches for query autocompletion." In SIGIR, pp. 775-778. 2015.
Manish Gupta (gmanish@microsoft.com)
22
DL for QAC, ECIR 23
Implicit Negative Feedback from Previous Prefixes in same conversation
Zhang, Aston, Amit Goyal, Weize Kong, Hongbo Deng, Anlei Dong, Yi Chang, Carl A. Gunter, and Jiawei Han. "adaqac: Adaptive query auto-completion via implicit negative feedback." In SIGIR, pp. 143-152. 2015.
Manish Gupta (gmanish@microsoft.com)
23
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
24
DL for QAC, ECIR 23
Convolutional latent semantic model for rare prefixes
Manish Gupta (gmanish@microsoft.com)
25
DL for QAC, ECIR 23
CLSM architecture
Manish Gupta (gmanish@microsoft.com)
26
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
27
DL for QAC, ECIR 23
Efficient Generation and Ranking for Neural QAC
Wang, Sida, Weiwei Guo, Huiji Gao, and Bo Long. "Efficient Neural Query Auto Completion." In CIKM, pp. 2797-2804. 2020.
Manish Gupta (gmanish@microsoft.com)
28
DL for QAC, ECIR 23
Our neural ranking model architecture. On top of it is a Learning-To-Rank layer that takes in multiple candidate scores. The input query has a special token “<sos>"; the probability of "research scientist <eos>" is computed based on LSTM hidden states.
Efficient Generation and Ranking for Neural QAC
Wang, Sida, Weiwei Guo, Huiji Gao, and Bo Long. "Efficient Neural Query Auto Completion." In CIKM, pp. 2797-2804. 2020.
Manish Gupta (gmanish@microsoft.com)
29
DL for QAC, ECIR 23
The average time cost of ranking a candidate list with 10 candidates is measured for each model. The average number of words in candidates is 3.20. The hidden vector size and embedding size of LM is 100 and the LSTM layer number is 1.
LSTMEmbed: The final hidden state vector from LSTM is used as the semantic representation of the sequence.
Agenda
Manish Gupta (gmanish@microsoft.com)
30
DL for QAC, ECIR 23
Personalization for QAC
Manish Gupta (gmanish@microsoft.com)
31
DL for QAC, ECIR 23
Using short-term/long-term user history, location, other signals
the great gatsby fitzgerald
the great gatsby book
the great influenza
the great alone book
the great gatsby film 2013
Search/Click History:
tender is the night
f. scott fitzgerald
this side of paradise
the silent patient
the great gatsby book
the great gatsby film 2013
the great gatsby trailer
the great gatsby film 1974
the greatest showman
the great wall film 2016
Search/Click History:
the wolf of wall street
the revenant
inception
gangs of new york
pain & gain
the great wall of china
the great gatsby film 2013
the great gatsby trailer
the great gatsby film 1974
the greatest showman
Search/Click History:
the wolf of wall street
the revenant
inception
gangs of new york
pain & gain
beijing travel advisory
forbidden city
Session history
Long-term
Agenda
Manish Gupta (gmanish@microsoft.com)
32
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
33
DL for QAC, ECIR 23
Motivation for personalization
Manish Gupta (gmanish@microsoft.com)
34
DL for QAC, ECIR 23
(Top) The likelihood of instagram and imdb in queries submitted by different demographics according to Yahoo! Clues. (Bottom) The likelihood of instagram and imdb in queries submitted by the logged-in users of Bing.
Features for personalizing auto-completion
Manish Gupta (gmanish@microsoft.com)
35
DL for QAC, ECIR 23
LTR framework for personalization
Manish Gupta (gmanish@microsoft.com)
36
DL for QAC, ECIR 23
The biggest movers in personalized autocompletion rankings when the ranker is trained by age features. Each column includes the candidates that were boosted most frequently in the personalized auto-completion rankings for users of the specified age groups.
Location for personalization
Manish Gupta (gmanish@microsoft.com)
37
DL for QAC, ECIR 23
The top movers in each region. These are queries that their average positions in rankings with and without personalization differ the most in each region. The regions are specified by collapsing the first zip-code digits and the users in each region are grouped accordingly. Each map shows the distribution of query popularity across different US states according to Google Trends, and the colors range between light blue (rare) and dark blue (popular).
Agenda
Manish Gupta (gmanish@microsoft.com)
38
DL for QAC, ECIR 23
RNNs for personalization
Song, Jun, Jun Xiao, Fei Wu, Haishan Wu, Tong Zhang, Zhongfei Mark Zhang, and Wenwu Zhu. "Hierarchical contextual attention recurrent neural network for map query suggestion." IEEE TKDE 29, no. 9 (2017): 1888-1901.
Manish Gupta (gmanish@microsoft.com)
39
DL for QAC, ECIR 23
Attend, Copy, Generate (ACG)
Manish Gupta (gmanish@microsoft.com)
40
DL for QAC, ECIR 23
Attend, Copy, Generate (ACG)
Manish Gupta (gmanish@microsoft.com)
41
DL for QAC, ECIR 23
Example of generating a suggestion query given the previous queries in the session. The suggestion query is generated during three time steps. The heatmap indicates the attention, red for query-level attention and blue for word-level attention. The pie chart shows if the network decides to copy or to generate.
Manish Gupta (gmanish@microsoft.com)
42
DL for QAC, ECIR 23
Evaluation
Manish Gupta (gmanish@microsoft.com)
43
DL for QAC, ECIR 23
Comparisons
Manish Gupta (gmanish@microsoft.com)
44
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
45
DL for QAC, ECIR 23
Personalized neural Language Model
Manish Gupta (gmanish@microsoft.com)
46
DL for QAC, ECIR 23
Personalized neural Language Model
Manish Gupta (gmanish@microsoft.com)
47
DL for QAC, ECIR 23
Diverse beam search
Vijayakumar, Ashwin K., Michael Cogswell, Ramprasath R. Selvaraju, Qing Sun, Stefan Lee, David Crandall, and Dhruv Batra. "Diverse beam search: Decoding diverse solutions from neural sequence models." arXiv preprint arXiv:1610.02424 (2016).
Manish Gupta (gmanish@microsoft.com)
48
DL for QAC, ECIR 23
Manish Gupta (gmanish@microsoft.com)
49
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
50
DL for QAC, ECIR 23
Multi-view Multi-task Attentive framework for Personalized QAC
Manish Gupta (gmanish@microsoft.com)
51
DL for QAC, ECIR 23
Examples of query candidates recommended by different models for the same prefix and history behavior.
Manish Gupta (gmanish@microsoft.com)
52
DL for QAC, ECIR 23
Ablation Study of query generation models.
Agenda
Manish Gupta (gmanish@microsoft.com)
53
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
54
DL for QAC, ECIR 23
Inappropriate query suggestion detection
Yenala, Harish, Manoj Chinnakotla, and Jay Goyal. "Convolutional Bi-directional LSTM for detecting inappropriate query suggestions in web search." In PAKDD, pp. 3-16. Springer, Cham, 2017.
Manish Gupta (gmanish@microsoft.com)
55
DL for QAC, ECIR 23
CONV+BiLSTMs for Inappropriate query suggestion detection
Yenala, Harish, Manoj Chinnakotla, and Jay Goyal. "Convolutional Bi-directional LSTM for detecting inappropriate query suggestions in web search." In PAKDD, pp. 3-16. Springer, Cham, 2017.
Manish Gupta (gmanish@microsoft.com)
56
DL for QAC, ECIR 23
Results
Yenala, Harish, Manoj Chinnakotla, and Jay Goyal. "Convolutional Bi-directional LSTM for detecting inappropriate query suggestions in web search." In PAKDD, pp. 3-16. Springer, Cham, 2017.
Manish Gupta (gmanish@microsoft.com)
57
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
58
DL for QAC, ECIR 23
Spell correction using soft-masked BERT
Zhang, Shaohua, Haoran Huang, Jicong Liu, and Hang Li. "Spelling Error Correction with Soft-Masked BERT." In ACL, pp. 882-890. 2020.
Manish Gupta (gmanish@microsoft.com)
59
DL for QAC, ECIR 23
Soft-masked BERT
Zhang, Shaohua, Haoran Huang, Jicong Liu, and Hang Li. "Spelling Error Correction with Soft-Masked BERT." In ACL, pp. 882-890. 2020.
Manish Gupta (gmanish@microsoft.com)
60
DL for QAC, ECIR 23
Detection loss
Correction loss
Agenda
Manish Gupta (gmanish@microsoft.com)
61
DL for QAC, ECIR 23
Need for online spell correction
Duan, Huizhong, and Bo-June Hsu. "Online spelling correction for query completion." In Proceedings of the 20th international conference on World wide web, pp. 117-126. 2011.
Manish Gupta (gmanish@microsoft.com)
62
DL for QAC, ECIR 23
Trie with online spell correction
Duan, Huizhong, and Bo-June Hsu. "Online spelling correction for query completion." In Proceedings of the 20th international conference on World wide web, pp. 117-126. 2011.
Manish Gupta (gmanish@microsoft.com)
63
DL for QAC, ECIR 23
From | To | Cost |
a | b | 5000 |
a | c | 5000 |
a | d | 6000 |
Conversion Table
Exploration: Path are explored with exact matching or paying some conversion cost
Example for Prefix “a”: the explored trie is highlighted in yellow and this is the resulting priority queue:
a 3000 | b 1000+5000 | d 1000 + 6000 | c 5000 + 5000 |
Learning conversion rules
Duan, Huizhong, and Bo-June Hsu. "Online spelling correction for query completion." In Proceedings of the 20th international conference on World wide web, pp. 117-126. 2011.
Manish Gupta (gmanish@microsoft.com)
64
DL for QAC, ECIR 23
Error correction with char RNNs
Wang, Po-Wei, Huan Zhang, Vijai Mohan, Inderjit S. Dhillon, and J. Zico Kolter. "Realtime query completion via deep language models." In eCOM@ SIGIR. 2018.
Manish Gupta (gmanish@microsoft.com)
65
DL for QAC, ECIR 23
Completion vs Error Correction
Wang, Po-Wei, Huan Zhang, Vijai Mohan, Inderjit S. Dhillon, and J. Zico Kolter. "Realtime query completion via deep language models." In eCOM@ SIGIR. 2018.
Manish Gupta (gmanish@microsoft.com)
66
DL for QAC, ECIR 23
General Beam Search
Edit Distance v.s. Completion Distance
Beam Search with Edit Distance
Wang, Po-Wei, Huan Zhang, Vijai Mohan, Inderjit S. Dhillon, and J. Zico Kolter. "Realtime query completion via deep language models." In eCOM@ SIGIR. 2018.
Manish Gupta (gmanish@microsoft.com)
67
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
68
DL for QAC, ECIR 23
NLG for QAC
Manish Gupta (gmanish@microsoft.com)
69
DL for QAC, ECIR 23
Query Blazer: NLG without deep learning
Kang, Young Mo, Wenhao Liu, and Yingbo Zhou. "QueryBlazer: Efficient Query Autocompletion Framework." In Proceedings of the 14th ACM International Conference on Web Search and Data Mining, pp. 1020-1028. 2021.
Manish Gupta (gmanish@microsoft.com)
70
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
71
DL for QAC, ECIR 23
Character-level Neural Language Model.
Manish Gupta (gmanish@microsoft.com)
72
DL for QAC, ECIR 23
Top suggested queries by char-LM. Phrases such as “afraid of the dead” and “afraid of the dog” and all prefixes do not exist in the data. Note that there is also a low-quality suggestion “when is a good time to buy a lyrics.
Architecture of our language model for an example query where ‘\n’ indicates the end of the query. Green cells contain one-hot encoded vectors of characters, blue cells contain character-embedded vectors, and red cells contain word-embedded vectors. <INC> means incomplete word token.
Character-level Neural Language Model.
Manish Gupta (gmanish@microsoft.com)
73
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
74
DL for QAC, ECIR 23
Subword Language Model for QAC
Kim, Gyuwan. "Subword language model for query auto-completion." arXiv preprint arXiv:1909.00599 (2019).
Manish Gupta (gmanish@microsoft.com)
75
DL for QAC, ECIR 23
Subword Language Model for QAC
Kim, Gyuwan. "Subword language model for query auto-completion." arXiv preprint arXiv:1909.00599 (2019).
Manish Gupta (gmanish@microsoft.com)
76
DL for QAC, ECIR 23
Subword Language Model for QAC
Kim, Gyuwan. "Subword language model for query auto-completion." arXiv preprint arXiv:1909.00599 (2019).
Manish Gupta (gmanish@microsoft.com)
77
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
78
DL for QAC, ECIR 23
Hierarchical recurrent encoder-decoder (HRED)
Sordoni, Alessandro, Yoshua Bengio, Hossein Vahabi, Christina Lioma, Jakob Grue Simonsen, and Jian-Yun Nie. "A hierarchical recurrent encoder-decoder for generative context-aware query suggestion." In CIKM, pp. 553-562. 2015.
Manish Gupta (gmanish@microsoft.com)
79
DL for QAC, ECIR 23
The user types “cleveland gallery → lake erie art”. During training, the model encodes “cleveland gallery”, updates the session-level recurrent state and maximizes the probability of seeing “lake erie art”. The process is repeated for all queries in the session. During testing, a contextual suggestion is generated by encoding the previous queries, by updating the session-level recurrent states accordingly and by sampling a new query from the last obtained session-level recurrent state. Here, the generated contextual suggestion is “cleveland indian art”.
HRED with LambdaMART
Sordoni, Alessandro, Yoshua Bengio, Hossein Vahabi, Christina Lioma, Jakob Grue Simonsen, and Jian-Yun Nie. "A hierarchical recurrent encoder-decoder for generative context-aware query suggestion." In CIKM, pp. 553-562. 2015.
Manish Gupta (gmanish@microsoft.com)
80
DL for QAC, ECIR 23
Comparison of HRED with BaselineRanker and ADJ
Sordoni, Alessandro, Yoshua Bengio, Hossein Vahabi, Christina Lioma, Jakob Grue Simonsen, and Jian-Yun Nie. "A hierarchical recurrent encoder-decoder for generative context-aware query suggestion." In CIKM, pp. 553-562. 2015.
Manish Gupta (gmanish@microsoft.com)
81
DL for QAC, ECIR 23
RIN: Reformulation Inference Network for Context-Aware Query Suggestion
Manish Gupta (gmanish@microsoft.com)
82
DL for QAC, ECIR 23
RIN: Reformulation Inference Network for Context-Aware Query Suggestion
Manish Gupta (gmanish@microsoft.com)
83
DL for QAC, ECIR 23
Self attention
RIN: Reformulation Inference Network for Context-Aware Query Suggestion
Manish Gupta (gmanish@microsoft.com)
84
DL for QAC, ECIR 23
RIN: Reformulation Inference Network for Context-Aware Query Suggestion
Manish Gupta (gmanish@microsoft.com)
85
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
86
DL for QAC, ECIR 23
Next Phrase Prediction for QAC for Emails/Academic Writings
Lee, Dong-Ho, Zhiqiang Hu, and Roy Ka-Wei Lee. "Improving Text Auto-Completion with Next Phrase Prediction." arXiv preprint arXiv:2109.07067 (2021).
Manish Gupta (gmanish@microsoft.com)
87
DL for QAC, ECIR 23
GPT-2 can generate syntactically sound, and semantically general sentence from partial query. However, it still needs to be fine-tuned a lot to generate semantically expert domain (e.g. Computer Science) focused sentence.
NPP Results
Lee, Dong-Ho, Zhiqiang Hu, and Roy Ka-Wei Lee. "Improving Text Auto-Completion with Next Phrase Prediction." arXiv preprint arXiv:2109.07067 (2021).
Manish Gupta (gmanish@microsoft.com)
88
DL for QAC, ECIR 23
Agenda
Manish Gupta (gmanish@microsoft.com)
89
DL for QAC, ECIR 23
When Are Search Completion Suggestions Problematic?
Olteanu, Alexandra, Fernando Diaz, and Gabriella Kazai. "When Are Search Completion Suggestions Problematic?." Proceedings of the ACM on Human-Computer Interaction 4, no. CSCW2 (2020): 1-25.
Manish Gupta (gmanish@microsoft.com)
90
DL for QAC, ECIR 23
When Are Search Completion Suggestions Problematic?
Olteanu, Alexandra, Fernando Diaz, and Gabriella Kazai. "When Are Search Completion Suggestions Problematic?." Proceedings of the ACM on Human-Computer Interaction 4, no. CSCW2 (2020): 1-25.
Manish Gupta (gmanish@microsoft.com)
91
DL for QAC, ECIR 23
Problem Category | Working definitions (and sub-categories) | Keywords (p/s: query) |
Harmful speech |
|
|
Potentially illicit |
|
|
When Are Search Completion Suggestions Problematic?
Olteanu, Alexandra, Fernando Diaz, and Gabriella Kazai. "When Are Search Completion Suggestions Problematic?." Proceedings of the ACM on Human-Computer Interaction 4, no. CSCW2 (2020): 1-25.
Manish Gupta (gmanish@microsoft.com)
92
DL for QAC, ECIR 23
Problem Category | Working definitions (and sub-categories) | Keywords (p/s: query) |
Controversy, Misinformation, and Manipulation |
|
|
Stereotypes & Bias |
|
|
Adult queries |
|
|
Other types |
|
|
When Are Search Completion Suggestions Problematic?
Olteanu, Alexandra, Fernando Diaz, and Gabriella Kazai. "When Are Search Completion Suggestions Problematic?." Proceedings of the ACM on Human-Computer Interaction 4, no. CSCW2 (2020): 1-25.
Manish Gupta (gmanish@microsoft.com)
93
DL for QAC, ECIR 23
Target Category | Working definitions (and sub-categories) | Keywords (p/s: query) |
Individuals |
|
|
Groups |
|
|
Businesses |
|
|
Organizations |
|
|
When Are Search Completion Suggestions Problematic?
Olteanu, Alexandra, Fernando Diaz, and Gabriella Kazai. "When Are Search Completion Suggestions Problematic?." Proceedings of the ACM on Human-Computer Interaction 4, no. CSCW2 (2020): 1-25.
Manish Gupta (gmanish@microsoft.com)
94
DL for QAC, ECIR 23
Target Category | Working definitions (and sub-categories) | Keywords (p/s: query) |
Animals & objects |
|
|
Activities & ideas |
|
|
Other targets |
|
|
Generic, no target |
|
|
Agenda
Manish Gupta (gmanish@microsoft.com)
95
DL for QAC, ECIR 23
Components in Query Auto Completion systems
Manish Gupta (gmanish@microsoft.com)
96
DL for QAC, ECIR 23
Ranking
Manish Gupta (gmanish@microsoft.com)
97
DL for QAC, ECIR 23
Personalization
Manish Gupta (gmanish@microsoft.com)
98
DL for QAC, ECIR 23
Handling defective suggestions and prefixes
Manish Gupta (gmanish@microsoft.com)
99
DL for QAC, ECIR 23
Natural Language Generation
Manish Gupta (gmanish@microsoft.com)
100
DL for QAC, ECIR 23
Extreme Multi-label Classification (XC/XMR) for QAC
Manish Gupta (gmanish@microsoft.com)
101
DL for QAC, ECIR 23
Yadav, Nishant, Rajat Sen, Daniel N. Hill, Arya Mazumdar, and Inderjit S. Dhillon. "Session-aware query auto-completion using extreme multi-label ranking." In PKDD, pp. 3835-3844. 2021.
Personalized NLG
Manish Gupta (gmanish@microsoft.com)
102
DL for QAC, ECIR 23
References
[1] Mostafa Dehghani, Sascha Rothe, Enrique Alfonseca, and Pascal Fleury. Learning to attend, copy, and generate for session-based query suggestion. In Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, pages 1747–1756, 2017.
[2] Giovanni Di Santo, Richard McCreadie, Craig Macdonald, and Iadh Ounis. Comparing approaches for query autocompletion. In Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 775–778, 2015.
[3] Huizhong Duan and Bo-June Hsu. Online spelling correction for query completion. In Proceedings of the 20th international conference on World wide web, pages 117–126, 2011.
[4] Nicolas Fiorini and Zhiyong Lu. Personalized neural language models for real-world query auto completion. In Proceedings of NAACL-HLT, pages 208–215, 2018.
[5] Jyun-Yu Jiang, Yen-Yu Ke, Pao-Yu Chien, and Pu-Jen Cheng. Learning user reformulation behavior for query auto-completion. In Proceedings of the 37th international ACM SIGIR conference on Research & development in information retrieval, pages 445–454, 2014.
[6] Jyun-Yu Jiang and Wei Wang. Rin: Reformulation inference network for context-aware query suggestion. In Proceedings of the 27th ACM International Conference on Information and Knowledge Management, pages 197–206, 2018.
[7] Gyuwan Kim. Subword language model for query autocompletion. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 5022– 5032, 2019.
[8] Dong-Ho Lee, Zhiqiang Hu, and Roy Ka-Wei Lee. Improving text auto-completion with next phrase prediction. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 4434–4438, 2021.
[9] Bhaskar Mitra and Nick Craswell. Query autocompletion for rare prefixes. In Proceedings of the 24th ACM international on conference on information and knowledge management, pages 1755–1758, 2015.
[10] Agnes Mustar, Sylvain Lamprier, and Benjamin Piwowarski. Using bert and bart for query suggestion. In Joint Conference of the Information Retrieval Communities in Europe, volume 2621. CEUR-WS. org, 2020.
Manish Gupta (gmanish@microsoft.com)
103
DL for QAC, ECIR 23
References
[11] Alexandra Olteanu, Fernando Diaz, and Gabriella Kazai. When are search completion suggestions problematic? Proceedings of the ACM on Human-Computer Interaction, 4(CSCW2):1–25, 2020.
[12] Dae Hoon Park and Rikio Chiba. A neural language model for query auto-completion. In Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 1189– 1192, 2017.
[13] Milad Shokouhi. Learning to personalize query autocompletion. In Proceedings of the 36th international ACM SIGIR conference on Research and development in information retrieval, pages 103–112, 2013.
[14] Jun Song, Jun Xiao, Fei Wu, Haishan Wu, Tong Zhang, Zhongfei Mark Zhang, and Wenwu Zhu. Hierarchical contextual attention recurrent neural network for map query suggestion. IEEE Transactions on Knowledge and Data Engineering, 29(9):1888–1901, 2017.
[15] Alessandro Sordoni, Yoshua Bengio, Hossein Vahabi, Christina Lioma, Jakob Grue Simonsen, and Jian-Yun Nie. A hierarchical recurrent encoder-decoder for generative context-aware query suggestion. In proceedings of the 24th ACM international on conference on information and knowledge management, pages 553–562, 2015.
[16] Po-Wei Wang, Huan Zhang, Vijai Mohan, Inderjit S Dhillon, and J Zico Kolter. Realtime query completion via deep language models. In eCOM@ SIGIR, 2018.
[17] Sida Wang, Weiwei Guo, Huiji Gao, and Bo Long. Efficient neural query auto completion. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management, pages 2797–2804, 2020.
[18] Harish Yenala, Manoj Chinnakotla, and Jay Goyal. Convolutional bi-directional lstm for detecting inappropriate query suggestions in web search. In Pacific-Asia Conference on Knowledge Discovery and Data Mining, pages 3–16. Springer, 2017.
[19] Di Yin, Jiwei Tan, Zhe Zhang, Hongbo Deng, Shujian Huang, and Jiajun Chen. Learning to generate personalized query auto-completions via a multi-view multitask attentive approach. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 2998–3007, 2020.
[20] Shaohua Zhang, Haoran Huang, Jicong Liu, and Hang Li. Spelling error correction with soft-masked bert. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 882–890, 2020.
Manish Gupta (gmanish@microsoft.com)
104
DL for QAC, ECIR 23
Thanks!
Manish Gupta (gmanish@microsoft.com)
105
DL for QAC, ECIR 23