Prepoznavanje pozicijskih motivov v šahovskih pozicijah

Plut, Matic (author), Guid, Matej (mentor)

V diplomski nalogi smo najprej pripravili program za samodejno prepoznavanje pozicijskih motivov. Usmerili smo se v prepoznavanje kmečkih struktur, kot so osamljeni kmetje, prosti kmetje, podvojeni kmetje, zaostali kmetje, kmečke verige in kmečki otoki. Program smo implementirali v programskem jeziku Python in s pomočjo knjižnice python-chess, ki omogoča enostavno uporabo bitnih šahovnic. Razviti program bi lahko uporabili denimo za samodejno komentiranje šahovskih partij. V nadaljevanju smo naslovili problem iskanja vsebinsko podobnih šahovskih pozicij. Samodejno prepoznavanje kmečkih struktur smo kot komponento vključili v program, ki je sposoben poiskati vsebinsko podobne šahovske pozicije, pri tem pa upošteva tudi lastnosti kmečkih struktur na šahovnici. Uporabili smo tehniko iskanja po besedilnih datotekah, pri čemer nam je bila v pomoč odprtokodna knjižnica Apache Lucene. Razvoj je delno potekal v okolju Jupyter Notebook, ki med drugim omogoča tudi enostavno prikazovanje šahovskih pozicij. Iskanje vsebinsko podobnih pozicij lahko uporabimo denimo pri samodejni pripravi šahovskih treningov, s čimer bi močno olajšali delo šahovskim trenerjem. Algoritem za iskanje podobnih pozicij bi lahko še izboljšali s pomočjo prepoznavanja dinamičnih, taktičnih motivov. Kompetentnost programa za ugotavljanje vsebinsko podobnih pozicij je bila potrjena s pomočjo ankete med šahisti.

• Language: Slovenian
• Keywords: umetna inteligenca, šah, pozicijski motivi, kmečke strukture, bitne šahovnice, podobne šahovske pozicije, poizvedba z vzorcem
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2018
• PID: 20.500.12556/RUL-104645
• Publication date in RUL: 09.10.2018

Secondary language

• Language: English
• Title: Recognition of positional motifs in chess positions

Abstract

In this diploma thesis, we first prepared a programme for automatic recognising of positional motifs. We focused on recognising of pawn structures, such as isolated pawns, passed pawns, doubled pawns, backward pawns, pawn chains, and pawn islands. The programme was implemented in the Python programming language with the help of the library python-chess that enables simple use of bitboards. The developed programme could be used for automatic annotation of chess games. Then, we addressed the problem of searching for chess positions with similar content. The automatic recognising of pawn structures was integrated into the programme as a component, wherein the programme is able to find chess positions with similar content, taking into account the characteristics of pawn structures on a chessboard as well. We used techniques for information retrieval from text documents, using an open-source library Apache Lucene. Partly, the development took place in the environment Jupyter Notebook, which enables simple visualization of chess positions. The search of positions with similar content can be used for automatic preparation of chess trainings, whereby the job of chess trainers can be made easier. We could improve the algorithm for searching similar positions by recognising dynamic, tactical motifs. The competence of the programme for discovering positions with similar content was confirmed by a survey carried out among chess players.

• Keywords: artificial intelligence, chess, positional motifs, pawn structures, bitboards, similar chess positions, query by example