PQPS: Prior-Art Query-Based Patent Summarizer Using RBM and Bi-LSTM

Table 3

Patent features for RBM-EPS.

---

Feature with description

TextRank (): this graph-based model scores the sentences by applying PageRank. It [32] computes the score based on the number of overlapping words between sentences.

---

Cue phrases: these phrases such as “in particular,” “in summary,” “as a result,” “as a consequence,” and so on are usually followed by important information, and so they are good indicators for estimating the sentence salience [69].

. Here, denotes the number of cue phrases in sentence and indicates the total number of cue phrases in the document.

---

Sentence semantic relatedness score (SSR): this score computes the relatedness between the search query and the phrases in a sentence with respect to smart device ontology using Lin’s measure [70]. Higher value on SSR score denotes they are more related and if there is no semantic match, SSR is assigned to 0.

represents the summation of Lin’s similarity between search query and phrase in sentence and denotes the total number of phrases in sentence . and are two ontologically related concepts, represents their most common ancestor, denotes the number of subsumers of a concept , and denotes the maximum number of nodes (concepts) in the ontology.

---