def __init__(self, dataDir, bigendian=False):
self.categorys = CharCategory.readCategorys(dataDir, bigendian)
fmis = FileMappedInputStream(dataDir + "/code2category", bigendian)
try:
self.char2id = fmis.getIntArray(fmis.size() // 4 // 2)
self.eqlMasks = fmis.getIntArray(fmis.size() // 4 // 2)
finally:
fmis.close()
def __init__(self, dataDir, bigendian=False, splitted=False):
self.trie = Searcher(dataDir + "/word2id", bigendian)
if splitted:
paths = sorted(glob.glob(dataDir + "/word.dat.*"))
self.data = util.getCharArrayMulti(paths, bigendian)
else:
self.data = util.getCharArray(dataDir + "/word.dat", bigendian)
self.indices = util.getIntArray(dataDir + "/word.ary.idx", bigendian)
fmis = FileMappedInputStream(dataDir + "/word.inf", bigendian)
try:
wordCount = fmis.size() // (4 + 2 + 2 + 2)
self.dataOffsets = fmis.getIntArray(wordCount)
""" dataOffsets[単語ID] = 単語の素性データの開始位置 """
self.leftIds = fmis.getShortArray(wordCount)
""" leftIds[単語ID] = 単語の左文脈ID """
self.rightIds = fmis.getShortArray(wordCount)
""" rightIds[単語ID] = 単語の右文脈ID """
self.costs = fmis.getShortArray(wordCount)
""" consts[単語ID] = 単語のコスト """
finally:
fmis.close()