論文地址：https://aclanthology.org/P19-1096.pdf
背景
情緒原因提取（ECE）旨在提取導(dǎo)致文本中情緒表達(dá)的潛在原因炫惩。
Figure 1 displays an example from this corpus, There are five clauses in a document.
The emotion "happy" is contained in the fourth clause. We denote this clause as emotion clause, which refers to a clause that contains emotions. It has two corresponding causes: "a policeman visited the old man with the lost money" in the second clause, and "told him that the thief was caught" in the third clause.
We denote them as cause clause, which refers to a clause that contains causes.
Take Figure 1 for example, given the annotation of emotion: "happy", the goal of ECE is to track the two corresponding cause clauses: "a policeman visited the old man with the lost money" and "and told him that the thief was caught". While in the ECPE task, the goal is to directly extract all pairs of emotion clause and cause clause, including ("The old man was very happy", "a policeman visited the old man with the lost money") and ("The old man was very happy", "and told him that the thief was caught"), without providing the emotion annotation "happy".
問題
1） 在ECE中蹋绽，在提取原因之前必須對(duì)情緒進(jìn)行注釋筋蓖，這極大地限制了其在現(xiàn)實(shí)世界場(chǎng)景中的應(yīng)用；
2） 先注釋情緒蚣抗，然后提取原因的方法忽略了情緒和原因是相互指示的這一事實(shí)瓮下。
解決辦法
我們提出了一種兩步方法來解決這一新的ECPE任務(wù)——
步驟1通過兩種多任務(wù)學(xué)習(xí)網(wǎng)絡(luò)將情緒-原因?qū)μ崛∪蝿?wù)轉(zhuǎn)換為兩個(gè)子任務(wù)（分別為情緒提取和原因提确砘怠），目的是提取一組情緒從句和一組原因從句迷捧。
The lower layer consists of a set of word-level Bi-LSTM modules, each of which corresponds to one clause, and accumulate the context information for each word of the clause. The hidden state of the jth word in the ith clause hi,j is obtained based on a bi-directional LSTM. Attention mechanism is then adopt to get a clause representation si.
The upper layer consists of two components: one for emotion extraction and another for cause extraction. Each component is a clause-level BiLSTM which receives the independent clause representations [s1, s2, ..., s|d|] obtained at the lower layer as inputs. The hidden states of two component Bi-LSTM, re i and rc i , can be viewed as the context-aware representation of clause ci, and finally feed to the softmax layer for emotion prediction and cause predication.
由于提供情緒可以幫助更好地發(fā)現(xiàn)原因笙蒙；了解原因也可能有助于更準(zhǔn)確地提取情緒手趣。受此啟發(fā)绿渣，我們進(jìn)一步提出了an interactive multi-task learning network中符，作為前一個(gè)網(wǎng)絡(luò)的增強(qiáng)版本淀散，以捕捉情緒和原因之間的相關(guān)性蚜锨。其結(jié)構(gòu)如圖3所示。
步驟2執(zhí)行情感原因配對(duì)和過濾亚再。我們通過將笛卡爾乘積應(yīng)用于情感集E和原因集C郭膛，將它們配對(duì)氛悬。這就產(chǎn)生了一組候選的情緒-原因?qū)ΑＷ詈笥?xùn)練一個(gè)濾波器來消除不包含因果關(guān)系的對(duì)如捅。
相關(guān)工作
Cheng等人（2017）專注于使用多用戶結(jié)構(gòu)對(duì)中國微博進(jìn)行原因檢測(cè)棍现。他們正式化了微博的兩個(gè)原因檢測(cè)任務(wù)（基于當(dāng)前子轉(zhuǎn)發(fā)的原因檢測(cè)和基于原始子轉(zhuǎn)發(fā)的理由檢測(cè)）镜遣，并引入了SVM和LSTM來處理它們己肮。
Chen等人（2018a）提出了一種分層卷積神經(jīng)網(wǎng)絡(luò)（Hier-CNN），該網(wǎng)絡(luò)使用子句級(jí)編碼器和子轉(zhuǎn)發(fā)級(jí)編碼器分別結(jié)合單詞上下文特征和基于事件的特征烈涮。
實(shí)驗(yàn)
數(shù)據(jù)集
[1]GUI L, WU D, XU R, 等. Event-driven emotion cause extraction with corpus construction[C/OL]//Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, Austin, Texas. 2016. http://dx.doi.org/10.18653/v1/d16-1170. DOI:10.18653/v1/d16-1170.
評(píng)價(jià)指標(biāo)
the precision, recall, and F1 score
貢獻(xiàn)
我們提出了一個(gè)新的任務(wù)：情緒-原因配對(duì)提取（ECPE）讶舰。它解決了傳統(tǒng)ECE任務(wù)在提取原因之前依賴于對(duì)情緒的注釋的缺點(diǎn)肋乍，并允許將情緒原因分析應(yīng)用于真實(shí)世界的場(chǎng)景
我們提出了一個(gè)兩步框架來解決ECPE任務(wù)，該任務(wù)首先執(zhí)行個(gè)人情緒提取和原因提取敷存，然后進(jìn)行情緒-原因配對(duì)和過濾觅闽。
