E3: Entailment-driven Extracting and Editing for Conversational Machine Reading

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 2310–2320 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 2310 E3: Entailment-driven Extracting and Editing for Conversational Machine Reading Victor Zhong University of Washington vzhong@cs.washington.edu Luke Zettlemoyer University of Washington lsz@cs.washington.edu Abstract Conversational machine reading systems help users answer high-level questions (e.g. deter- mine if they qualify for particular govern- ment benefits) when they do not know the ex- act rules by which the determination is made (e.g. whether they need certain income levels or veteran status). The key challenge is that these rules are only provided in the form of a procedural text (e.g. guidelines from govern- ment website) which the system must read to figure out what to ask the user. We present a new conversational machine reading model that jointly extracts a set of decision rules from the procedural text while reasoning about which are entailed by the conversational his- tory and which still need to be edited to create questions for the user. On the recently intro- duced ShARC conversational machine read- ing dataset, our Entailment-driven Extract and Edit network (E3) achieves a new state-of-the- art, outperforming existing systems as well as a new BERT-based baseline. In addition, by explicitly highlighting which information still needs to be gathered, E3 provides a more ex- plainable alternative to prior work. We release source code for our models and experiments at https://github.com/vzhong/e3. 1 Introduction In conversational machine reading (CMR), a sys- tem must help users answer high-level questions by participating in an information gathering dia- log. For example, in Figure 1 the system asks a series of questions to help the user decide if they need to pay tax on their pension. A key chal- lenge in CMR is that the rules by which the deci- sion is made are only provided in natural language (e.g. the rule text in Figure 1). At every step of the conversation, the system must read the rules text and reason about what has already been said in to formulate the best next question. # 4. Tax when you live abroad If you’re not a UK resident, you don’t usually pay UK tax on your pension. But you might have to pay tax in the country you live in. There are a few exceptions - for example, UK civil service pensions will always be taxed in the UK. I get my money from a business I have. We get our funding from a private bank. Rule text User scenario Do I need to pay UK tax on my pension? Initial user question Are you a UK resident? No Are you receiving UK civil service pensions? Previous question Previous user response Model output Figure 1: A conversational machine reading example. The model is given a rule text document, which con- tains a recipe of implicit rules (underlined) for answer- ing the initial user question. At the start of the conver- sation, the user presents a scenario …

Content truncated to 500 words