Integrates Wortschatz frequencies

.gitignore

@@ -3,4 +3,6 @@ __pycache__/
 .mypy_cache/
 *.py[cdo]
 *.egg-info/
-*.log
+*.log
+**/tars
+**/freq_tsvs

languages/wikipron/src/frequencies/grab_wortschatz_data.py

@@ -0,0 +1,69 @@
+#!/usr/bin/env python
+
+import json
+import logging
+import os
+import requests
+import tarfile
+import time
+
+WORTSCHATZ_DICT_PATH = "wortschatz_languages.json"
+
+
+# Downloads the Wortschatz tarballs, roughly 10 GB of data.
+def download(data_to_grab):
+    to_retry = {}
+    os.makedirs("tars", exist_ok=True)
+    for language in data_to_grab:
+        url = data_to_grab[language]["data_url"]
+        with requests.get(url, stream=True) as response:
+            target_path = url.split("/")[-1]
+            logging.info("Downloading: %s", target_path)
+            if response.status_code == 200:
+                with open(f"tars/{target_path}", "wb") as f:
+                    f.write(response.raw.read())
+            else:
+                logging.info(
+                    "Status code %s while downloading %s",
+                    response.status_code,
+                    target_path,
+                )
+                to_retry[language] = data_to_grab[language]
+        # 30 seconds appears to not be enough, 60-70 seconds works well
+        # but takes a long time.
+        time.sleep(45)
+    return to_retry
+
+
+# Unpacks word frequency TSVs of tarballs, roughly 1 GB of data.
+def unpack():
+    os.mkdir("freq_tsvs")
+    for tarball in os.listdir("tars"):
+        logging.info("Unpacking: %s", tarball)
+        with tarfile.open(name=f"tars/{tarball}", mode="r:gz") as tar_data:
+            for file_entry in tar_data:
+                if file_entry.name.endswith("words.txt"):
+                    # Removes inconsistent path in tarballs
+                    # so freq_tsvs has uniform contents.
+                    file_entry.name = os.path.basename(file_entry.name)
+                    tar_data.extract(file_entry, "freq_tsvs")
+
+
+def main():
+    with open(WORTSCHATZ_DICT_PATH, "r") as langs:
+        languages = json.load(langs)
+
+    # Hack for repeatedly attempting to download Wortschatz data
+    # as a way of getting around 404 response from their server.
+    langs_to_retry = download(languages)
+    while langs_to_retry:
+        langs_to_retry = download(langs_to_retry)
+
+    unpack()
+
+
+if __name__ == "__main__":
+    logging.basicConfig(
+        format="%(filename)s %(levelname)s: %(message)s", level="INFO"
+    )
+    main()

languages/wikipron/src/frequencies/merge.py

@@ -0,0 +1,87 @@
+#!/usr/bin/env python
+
+import csv
+import json
+import logging
+import os
+import tempfile
+
+from grab_wortschatz_data import WORTSCHATZ_DICT_PATH
+
+
+def rewrite_wikipron_tsv(
+    wiki_tsv_affix, transcription_level, frequencies_dict
+):
+    # Complete WikiPron TSV path.
+    file_to_target = wiki_tsv_affix + transcription_level
+    # Will try to overwrite phonetic and phonemic WikiPron TSVs
+    # for all Wortschatz languages. WikiPron may not have both a
+    # phonetic and phonemic TSV for all languages.
+    try:
+        # This is written to be run after remove_duplicates.sh
+        # and retain sorted order.
+        with open(file_to_target, "r") as wiki_file:
+            wiki_tsv = csv.reader(
+                wiki_file, delimiter="\t", quoting=csv.QUOTE_NONE
+            )
+            with tempfile.NamedTemporaryFile(
+                mode="w", dir="../../tsv", delete=False
+            ) as source:
+                for word, pron in wiki_tsv:
+                    # Check if WikiPron word is in Wortschatz frequencies
+                    # else set frequency to 0.
+                    if word in frequencies_dict:
+                        print(
+                            f"{word}\t{pron}\t{frequencies_dict[word]}",
+                            file=source,
+                        )
+                    else:
+                        print(f"{word}\t{pron}\t0", file=source)
+                temp_path = source.name
+        os.replace(temp_path, file_to_target)
+    except FileNotFoundError as err:
+        logging.info("File not found: %s", err)
+
+
+def main():
+    with open(WORTSCHATZ_DICT_PATH, "r") as langs:
+        languages = json.load(langs)
+
+    word_freq_dict = {}
+    transcription = ["_phonetic.tsv", "_phonemic.tsv"]
+
+    for freq_file in os.listdir("freq_tsvs"):
+        # For accessing correct language in wortschatz_languages.json.
+        file_to_match = freq_file.rsplit("-", 1)[0]
+        logging.info("Currently working on: %s", file_to_match)
+
+        with open(f"freq_tsvs/{freq_file}", "r") as tsv:
+            frequencies_tsv = csv.reader(
+                tsv, delimiter="\t", quoting=csv.QUOTE_NONE
+            )
+            for row in frequencies_tsv:
+                # Wortschatz TSVs are not uniformly formatted.
+                # Some have 3 columns, some have 4.
+                try:
+                    word = row[2].lower()
+                    freq = int(row[3])
+                except IndexError:
+                    word = row[1].lower()
+                    freq = int(row[2])
+                # Filter out numbers in Wortschatz data.
+                if str.isalpha(word):
+                    if word not in word_freq_dict:
+                        word_freq_dict[word] = freq
+                    else:
+                        word_freq_dict[word] = word_freq_dict[word] + freq
+
+        for wiki_tsv_path in languages[file_to_match]["path"]:
+            for level in transcription:
+                rewrite_wikipron_tsv(wiki_tsv_path, level, word_freq_dict)
+
+
+if __name__ == "__main__":
+    logging.basicConfig(
+        format="%(filename)s %(levelname)s: %(message)s", level="INFO"
+    )
+    main()