A Tool for Efficient Content Compilation

We build a tool to assist in content creation by mining the web for information relevant to a given topic. This tool imitates the process of essay writing by humans: searching for topics on the web, selecting content fragments from the found document, and then compiling these fragments to obtain a coherent text. The process of writing starts with automated building of a table of content by obtaining the list of key entities for the given topic extracted from web resources such as Wikipedia. Once a table of content is formed, each item forms a seed for web mining. The tool builds a full-featured structured Word document with table of content, section structure, images and captions and web references for all included text fragments. Two linguistic technologies are employed: for relevance verification, we use similarity computed as a tree similarity between parse trees for a seed and candidate text fragment. For text coherence, we use a measure of agreement between a given and consecutive paragraph by tree kernel learning of their discourse trees. The tool is available at http://animatronica.io/submit.html. 1 Introducing content compilation problem In the modern society, writing and creating content is one of the most frequent human activity. An army of content creators, from students to professional writers produce various kinds of documents for various audiences. Not all of these documents are expected to be innovative, break-through or extremely important. The target of the tool being proposed is assistance with routine document creation process (Fig. 1) where most information is available on the web and needs to be collected, integrated and properly referenced. A number of content generation software systems are available in specific business domains (Johnson 2016). Most of content generation software are template-based which limits their efficiency and volume of produced content (Hendrikx et al 2015). An interesting class of content generation system is based on verbalizing some numerical data. Also, content generation for computer game support turned out to be fruitful (Liapis et al 2013). Deep-learning – based generation of a sequence of words has a limited applicability for large scale content production industrial systems. The goal of this study is to build a content compilation assistance system that would meet the following criteria: • Produces high volume cohesive text on a given topic in a domain-independent manner; • Collects text fragments from the web and combines them to assist in research on a given topic, provide systematic references; • Combines text, image and video resources in the resultant document; • Suitable for producing a final report and manual editing by students, researchers in various fields in science, engineering, business and law. On the bottom-left of Fig. 1 we show the main problem that needs to be solved to build a document from fragments collected from the web. For given two fragments, we need to determine if one can reasonably follow another in a cohesive manner. W build a discourse representation for each fragment an learn this representation to classify a pair of consecutive paragraphs as cohesive or not.