Joint Learning with Global Inference for Comment Classification in Community Question Answering
23336Authors
nan
Abstract
This paper addresses the problem of comment classification in community Question Answer- ing. Following the state of the art, we ap- proach the task with a global inference pro- cess to exploit the information of all com- ments in the answer-thread in the form of a fully connected graph. Our contribution com- prises two novel joint learning models that are on-line and integrate inference within learn- ing. The first one jointly learns two node- and edge-level MaxEnt classifiers with stochastic gradient descent and integrates the inference step with loopy belief propagation. The sec- ond model is an instance of fully connected pairwise CRFs (FCCRF). The FCCRF model significantly outperforms all other approaches and yields the best results on the task to date. Crucial elements for its success are the global normalization and an Ising-like edge potential. 1
Keywords
Full Content
Proceedings of NAACL-HLT 2016, pages 703–713, San Diego, California, June 12-17, 2016. c⃝2016 Association for Computational Linguistics Joint Learning with Global Inference for Comment Classification in Community Question Answering Shafiq Joty, Llu´ıs M`arquez and Preslav Nakov ALT Research Group Qatar Computing Research Institute — HBKU, Qatar Foundation {sjoty,lmarquez,pnakov}@qf.org.qa Abstract This paper addresses the problem of comment classification in community Question Answer- ing. Following the state of the art, we ap- proach the task with a global inference pro- cess to exploit the information of all com- ments in the answer-thread in the form of a fully connected graph. Our contribution com- prises two novel joint learning models that are on-line and integrate inference within learn- ing. The first one jointly learns two node- and edge-level MaxEnt classifiers with stochastic gradient descent and integrates the inference step with loopy belief propagation. The sec- ond model is an instance of fully connected pairwise CRFs (FCCRF). The FCCRF model significantly outperforms all other approaches and yields the best results on the task to date. Crucial elements for its success are the global normalization and an Ising-like edge potential. 1 Introduction Online community fora have been gaining a lot of popularity in recent years. Many of them, such as Stack Exchange1, are quite open, allowing anybody to ask and anybody to answer a question, which makes them very valuable sources of information. Yet, this same democratic nature resulted in some questions accumulating a large number of answers, many of which are of low quality. While nowa- days online fora are typically searched using stan- dard search engines that index entire threads, this is not optimal, as it can be very time-consuming for a user to go through and make sense of a long thread. 1http://stackexchange.com/ Q: hello guys and gals..could anyone of u knows where to buy a good and originals RC helicopters and toy guns here in qatar..im longin for this toys but its nowhere to find.. thanks A1 Go to Doha city center you may get it at 4 floor. Local: Good, Human: Good A2 “Hobby Shop” in City center has these toys with orig- inal motors. They are super cool.. U will love that shop..and will definetly buy one :) Have fun :) Local: Good, Human: Good A3 IM selling all my rc nitro helicopters. call me at 5285113.. (1)TREX 600 new/ (1) TREX500 (1) SHUTTLERG (1) FUTABA ... [truncated] Local: Good, Human: Bad A4 Hobby Shop- City Centre Local: Bad, Human: Good A5 OMG!! :— Guns and helicopters??!! Local: Good, Human: Bad A6 Speed Marine- Salwa Road I think these guys r the best in town... Local: Good, Human: Good A7 City center, i’ve seen wonderful collection.. Its some wer besides the kids fun place.. Local: Bad, Human: Good A8 try the shop in city center. they have many RC toys for sale there. and for the toy guns, in your talking baout airsoft i think its prohibited here. good luck Local: Good, Human: Good Figure 1: Example answer-thread with …
Content truncated to 500 words
Labeled Text
This paper addresses the problem of comment classification in community Question Answering. Following the state of the art, we approach the task with a global inference process to exploit the information of all comments in the answer-thread in the form of a fully connected graph. Our contribution comprises two novel joint learning models that are on-line and integrate inference within learning. The first one jointly learns two node- and edge-level MaxEnt classifiers with stochastic gradient descent and integrates the inference step with loopy belief propagation. The second model is an instance of fully connected pairwise CRFs (FCCRF). The FCCRF model significantly outperforms all other approaches and yields the best results on the task to date. Crucial elements for its success are the global normalization and an Ising-like edge potential. 1
Publication Details
- Paper ID:
- 23336
- Year:
- 2016
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan