From 392c672a49ab1762a837eedb584149f583ea2c10 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Radim=20=C5=98eh=C5=AF=C5=99ek?= <radimrehurek@seznam.cz>
Date: Fri, 23 Jun 2017 23:41:21 +0900
Subject: [PATCH] wip: API design for fast_counter & phrases

---
 gensim/models/fast_counter.py | 123 ++++++++++++++++++++++++++++++++++
 gensim/models/phrases.py      |   6 --
 2 files changed, 123 insertions(+), 6 deletions(-)
 create mode 100644 gensim/models/fast_counter.py

diff --git a/gensim/models/fast_counter.py b/gensim/models/fast_counter.py
new file mode 100644
index 0000000000..431a3d39a2
--- /dev/null
+++ b/gensim/models/fast_counter.py
@@ -0,0 +1,123 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+#
+# Copyright (C) 2017 Radim Rehurek <me@radimrehurek.com>
+# Licensed under the GNU LGPL v2.1 - http://www.gnu.org/licenses/lgpl.html
+
+"""
+Fast & memory efficient counting of things (and n-grams of things).
+
+This module is designed to count *item* and *document* frequencies over large, streamed corpora (lazy iteration).
+
+"""
+
+from collections import Counter
+import logging
+
+from six import iterkeys, iteritems
+
+logger = logging.getLogger(__name__)
+
+
+def iter_ngrams(document, ngrams):
+    assert ngrams[0] <= ngrams[1]
+
+    for n in range(ngrams[0], ngrams[1] + 1):
+        for ngram in zip(*[document[i:] for i in range(n)]):
+            logger.debug("yielding ngram %r", ngram)
+            yield ngram
+
+def iter_gram1(document):
+    return iter_ngrams(document, (1, 1))
+
+def iter_gram2(document):
+    return iter_ngrams(document, (2, 2))
+
+def iter_gram12(document):
+    return iter_ngrams(document, (1, 2))
+
+
+class FastCounter(object):
+    """
+    Fast counting of item frequency and document frequency across large, streamed iterables.
+    """
+
+    def __init__(self, doc2items=iter_gram1, collect_df=False):
+        self.doc2items = doc2items
+        self.collect_df = collect_df
+
+        self.item_counts = Counter()  # TODO replace by some GIL-free low-level struct
+        self.doc_counts = Counter()  # TODO replace by some GIL-free low-level struct
+
+    def hash(self, key):
+        return hash(key)
+
+    def update(self, documents):
+        """
+        Update the relevant ngram counters from the iterable `documents`.
+
+        If the memory structures get too large, clip them (then the internal counts may be only approximate).
+        """
+        for document in documents:
+            # TODO: release GIL, so we can run update() in parallel threads.
+            # Or maybe not needed, if we create multiple FastCounters from multiple input streams using
+            # multiprocessing, and only .merge() them at the end.
+            item_cnts = Counter(self.hash(ngram) for ngram in self.doc2items(document))
+            self.item_counts.update(item_cnts)
+            if self.collect_df:
+                # increment by 1 per unique key ("document frequency")
+                self.doc_counts.update(iterkeys(item_cnts))
+
+            # self.prune_vocab()
+
+        return self  # for easier chaining
+
+    def prune_vocab(self):
+        # Trim data structures to fit in memory, if too large.
+        # Or use a fixed-size data structure to start with (hyperloglog?)
+        raise NotImplementedError
+
+    def get(self, key, default=None):
+        """Return the item frequency of `key` (or `default` if key not present)."""
+        return self.item_counts.get(self.hash(key), default)
+
+    def merge(self, other):
+        """
+        Merge counts from other into self, in-place.
+        """
+        # rare operation, no need to optimize too much
+        raise NotImplementedError
+
+    def __len__(self):
+        return len(self.item_counts)
+
+    def __str__(self):
+        return "%s<%i items>" % (self.__class__.__name__, len(self))
+
+
+class Phrases(object):
+    def __init__(self, min_count=5, threshold=10.0, max_vocab_size=40000000):
+        self.threshold = threshold
+        self.min_count = min_count
+        self.max_vocab_size = max_vocab_size
+        self.counter = FastCounter(iter_gram12)
+
+    def add_documents(self, documents):
+        self.counter.update(documents)
+
+        return self  # for easier chaining
+
+    def export_phrases(self, document):
+        """
+        Yield all collocations (pairs of adjacent closely related tokens) from the
+        input `document`, as 2-tuples `(score, bigram)`.
+        """
+        if not self.counter:
+            return
+        norm = 1.0 * len(self.counter)
+        for bigram in iter_gram2(document):
+            pa, pb, pab = self.counter.get((bigram[0],)), self.counter.get((bigram[1],)), self.counter.get(bigram, 0)
+            if pa and pb:
+                score = norm / pa / pb * (pab - self.min_count)
+                if score > self.threshold:
+                    yield score, bigram
diff --git a/gensim/models/phrases.py b/gensim/models/phrases.py
index be735b865a..65d9b93b02 100644
--- a/gensim/models/phrases.py
+++ b/gensim/models/phrases.py
@@ -134,12 +134,6 @@ def __init__(self, sentences=None, min_count=5, threshold=10.0,
         should be a byte string (e.g. b'_').
 
         """
-        if min_count <= 0:
-            raise ValueError("min_count should be at least 1")
-
-        if threshold <= 0:
-            raise ValueError("threshold should be positive")
-
         self.min_count = min_count
         self.threshold = threshold
         self.max_vocab_size = max_vocab_size