예제 #1
0
    def test_minhash_sketch_similarity(self):
        """Signatures should be consistent"""
        sh = MinHashSketchSignature(64)

        sig1 = sh.get_signature("abracadabra")
        sig2 = sh.get_signature("abracadabra")
        dist = hamming(sig1, sig2)
        self.assertEqual(dist, 0)

        sig1 = sh.get_signature("abracadabra")
        sig2 = sh.get_signature("arbcd")
        dist = hamming(sig1, sig2)
        self.assertEqual(dist, 0)

        sig1 = sh.get_signature("abracadabra")
        sig2 = sh.get_signature("")
        dist = hamming(sig1, sig2)
        self.assertEqual(dist, 26)
예제 #2
0
    def __init__(self,
                 cfg,
                 content_filter=None,
                 trace_every=0,
                 get_body=None,
                 get_label=None,
                 get_prefix=None,
                 min_support=None,
                 seed=0,
                 normalizer=None,
                 tokenizer=None):
        """Read configuration"""
        self.cfg = cfg
        self._get_body = get_body
        self._get_label = get_label
        self._get_prefix = get_prefix

        self.trace_every = trace_every

        # Set options
        self.content_filter = content_filter
        self.min_support = cfg[
            'min_support'] if min_support is None else min_support

        # normalizer and tokenizer
        self.normalizer = get_default_normalizer(
            **cfg.get('preprocessor', {}).get('normalizer', {})) \
            if normalizer is None else normalizer
        self.tokenizer = get_default_tokenizer(
        ) if tokenizer is None else tokenizer

        # Configure minhash signer
        sig_width = cfg['sig_width']
        lsh_hasher = LSHC(width=sig_width, **cfg['lsh_options'])
        self.signer = MinHashSignature(sig_width,
                                       lsh_hasher=lsh_hasher,
                                       kmin=cfg['kmin'])

        # Configure shingler
        cfg_key_shingle = cfg['shingler']
        self.shingler = get_default_shingler(**cfg_key_shingle)

        # Configure sketch comparison algorithm
        cfg_sketch = cfg['sketch']
        self.sketch_enabled = cfg_sketch['enabled']
        self.sketch_dist_fn = None
        self.max_dist = None
        if self.sketch_enabled:
            algorithm_name = cfg_sketch['algorithm']
            try:
                sketch_algorithm = getattr(SketchModel, algorithm_name)
            except AttributeError:
                raise RuntimeError("Unknown sketch model specified: '%s'" %
                                   algorithm_name)
            self.sketch_bits = cfg_sketch['size']
            cfg_sketch_shingler = cfg_sketch['shingler']
            if not cfg_sketch_shingler['enabled']:
                # if sketch shingler is disabled, we also disable signer
                # as we will use default signer
                self.sketch_shingler = None
                self.sketch_signer = None
            elif sketch_algorithm == SketchModel.simhash:
                del cfg_sketch_shingler['enabled']
                self.sketch_shingler = Shingler(**cfg_sketch_shingler)
                self.sketch_signer = SimHashSignature(self.sketch_bits,
                                                      seed=seed)
            elif sketch_algorithm == SketchModel.minhash:
                del cfg_sketch_shingler['enabled']
                self.sketch_shingler = Shingler(**cfg_sketch_shingler)
                self.sketch_signer = MinHashSketchSignature(self.sketch_bits,
                                                            seed=seed)

            self.sketch_shingler._tokenizer = None
            self.sketch_shingler._normalizer = None

            self.max_dist = \
                int(floor(self.sketch_bits *
                          (1.0 - float(cfg_sketch['resemblance']))))
            self.sketch_dist_fn = hamming
            self.sketch_operator = OPERATOR_MAP[cfg_sketch.get(
                'operator', 'and')]
        self.cluster_builder = Cluster(sketch_dist_fn=self.sketch_dist_fn,
                                       max_dist=self.max_dist,
                                       min_support=self.min_support,
                                       sketch_operator=self.sketch_operator)