Mapping Natural Language Instructions to Mobile UI Action Sequences

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 789–799 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 789 Interactive Construction of User-Centric Dictionary for Text Analytics Ryosuke Kohita kohi@ibm.com Issei Yoshida IBM Research Hiroshi Kanayama {issei,hkana,nasukawa}@jp.ibm.com Tetsuya Nasukawa Abstract We propose a methodology to construct a term dictionary for text analytics through an inter- active process between a human and a ma- chine. The interactive approach helps the cre- ation of flexible dictionaries with precise gran- ularity required in text analysis. This paper introduces the first formulation of interactive dictionary construction to address this issue. To optimize the interaction, we propose a new algorithm that effectively captures an analyst’s intention starting from only a small number of sample terms. Along with the algorithm, we also design an automatic evaluation frame- work that provides a systematic assessment of any interactive method for the dictionary cre- ation task. Experiments using real scenario based corpora and dictionaries show that our algorithm outperforms baseline methods, and works even with a small number of interac- tions. Also, we provide our dataset for future studies1. 1 Introduction Since the emergence of practical interests in text analytics that finds insights from massive docu- ments (Nasukawa and Nagano, 2001), there are several requirements for enhancing valuable dis- coveries. The one critical issue we tackle in this paper is an effective construction of a term dic- tionary (Godbole et al., 2010). The term dic- tionary, which is an arbitrary set of terms, is used in text analytics to represent interesting anal- ysis perspectives (Nasukawa and Nagano, 2001; Nasukawa, 2009); for example, dictionaries of “product names” and “evaluative description” are required for mining customer reputations about products. The motivation of this paper is how to reduce the human workload for the dictionary con- 1https://github.com/kohilin/ IDC-evalset.git Evaluative description Functional description Appearance description Yellow Dirty Waterproof Portable Sturdy User-Friendly Stretchy Superior Formal Traditional For pregnant Metalic Bad Studless Synonym of U.S.A Medicine name U.S., U.S.A., America, … Aspirin, Opdivo, Tylenol, … Typical dictionary in previous studies Beautiful Flexible and fine-grained dictionary in this work Nice Figure 1: Typical dictionaries in previous works (up- per) and fine-grained dictionaries in this work (lower) struction as much as possible. To this end, we es- tablish a methodology of interactive dictionary construction that incrementally captures an ana- lyst’s intention starting from a small number of sample terms and enables him/her to effortlessly expand terms in the intended dictionary through suggestions by a machine. The term dictionary for text analytics is expen- sive to be constructed because we need to fo- cus more on terms with flexible granularity for in-depth analysis (Takeuchi et al., 2009; Godbole et al., 2010; Mostafa, 2013). For instance, if the analyst wants to examine product evaluation from both its function and appearance, he/she then needs to separately create those dictionaries whose boundaries are vague and overlapped (Figure 1). In short, we need to group any terms the analyst wants together depending on documents and the …

Content truncated to 500 words