HU TN Fixes issue NVIDIA#166 (NVIDIA#184)

zoobereq · pre-commit-ci[bot] · web-flow · commit d7da1ce5f80a · 2024-07-18T15:47:42.000-04:00
* Fixes issue NVIDIA#166 Signed-off-by: Simon Zuberek <szuberek@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Implements aliases for common string literals Signed-off-by: Simon Zuberek <szuberek@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fixes the period variable Signed-off-by: Simon Zuberek <szuberek@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Simon Zuberek <szuberek@nvidia.com> Co-authored-by: Simon Zuberek <szuberek@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
diff --git a/Jenkinsfile b/Jenkinsfile
@@ -17,7 +17,7 @@ pipeline {
     ES_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/10-27-23-0'
     ES_EN_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-13-23-2'
     FR_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-04-24-0'
-    HU_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-08-23-0'
+    HU_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/07-16-24-0'
     PT_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-08-23-0'
     RU_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-08-23-0'
     VI_TN_CACHE='/home/jenkinsci/TestData/text_norm/ci/grammars/06-08-23-0'
diff --git a/nemo_text_processing/text_normalization/hu/taggers/electronic.py b/nemo_text_processing/text_normalization/hu/taggers/electronic.py
@@ -11,14 +11,29 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
+
 import pynini
 from pynini.lib import pynutil
 
-from nemo_text_processing.text_normalization.en.graph_utils import NEMO_ALPHA, NEMO_DIGIT, GraphFst, insert_space
-from nemo_text_processing.text_normalization.es.utils import get_abs_path, load_labels
-
-common_domains = [x[0] for x in load_labels(get_abs_path("data/electronic/domain.tsv"))]
-symbols = [x[0] for x in load_labels(get_abs_path("data/electronic/symbols.tsv"))]
+from nemo_text_processing.text_normalization.de.utils import get_abs_path, load_labels
+from nemo_text_processing.text_normalization.en.graph_utils import (
+    NEMO_ALPHA,
+    NEMO_DIGIT,
+    NEMO_SPACE,
+    GraphFst,
+    at,
+    colon,
+    domain_string,
+    double_quotes,
+    double_slash,
+    http,
+    https,
+    period,
+    protocol_string,
+    username_string,
+    www,
+)
 
 
 class ElectronicFst(GraphFst):
@@ -35,45 +50,65 @@ class ElectronicFst(GraphFst):
     def __init__(self, deterministic: bool = True):
         super().__init__(name="electronic", kind="classify", deterministic=deterministic)
 
-        dot = pynini.accep(".")
-        accepted_common_domains = pynini.union(*common_domains)
-        accepted_symbols = pynini.union(*symbols) - dot
-        accepted_characters = pynini.closure(NEMO_ALPHA | NEMO_DIGIT | accepted_symbols)
-        acceepted_characters_with_dot = pynini.closure(NEMO_ALPHA | NEMO_DIGIT | accepted_symbols | dot)
+        period_fst = pynini.accep(period)
+
+        symbols = [x[0] for x in load_labels(get_abs_path("data/electronic/symbols.tsv"))]
+        symbols = pynini.union(*symbols)
+        # all symbols
+        symbols_no_period = pynini.difference(symbols, period_fst)  # alphabet of accepted symbols excluding the '.'
+        accepted_characters = pynini.closure(
+            (NEMO_ALPHA | NEMO_DIGIT | symbols_no_period), 1
+        )  # alphabet of accepted chars excluding the '.'
+        all_characters = pynini.closure(
+            (NEMO_ALPHA | NEMO_DIGIT | symbols), 1
+        )  # alphabet of accepted chars including the '.'
+
+        # domains
+        domain = period_fst + accepted_characters
+        domain_graph = (
+            pynutil.insert(domain_string + colon + NEMO_SPACE + double_quotes)
+            + (accepted_characters + pynini.closure(domain, 1))
+            + pynutil.insert(double_quotes)
+        )
 
         # email
         username = (
-            pynutil.insert("username: \"")
-            + acceepted_characters_with_dot
-            + pynutil.insert("\"")
-            + pynini.cross('@', ' ')
+            pynutil.insert(username_string + colon + NEMO_SPACE + double_quotes)
+            + all_characters
+            + pynutil.insert(double_quotes)
+            + pynini.cross(at, NEMO_SPACE)
         )
-        domain_graph = accepted_characters + dot + accepted_characters
-        domain_graph = pynutil.insert("domain: \"") + domain_graph + pynutil.insert("\"")
-        domain_common_graph = (
-            pynutil.insert("domain: \"")
-            + accepted_characters
-            + accepted_common_domains
-            + pynini.closure((accepted_symbols | dot) + pynini.closure(accepted_characters, 1), 0, 1)
-            + pynutil.insert("\"")
+        email = username + domain_graph
+
+        # social media tags
+        tag = (
+            pynutil.delete(at)
+            + pynutil.insert(username_string + colon + NEMO_SPACE + double_quotes)
+            + (accepted_characters | (accepted_characters + pynini.closure(domain, 1)))
+            + pynutil.insert(double_quotes)
         )
-        graph = (username + domain_graph) | domain_common_graph
 
         # url
-        protocol_start = pynini.accep("https://") | pynini.accep("http://")
+        protocol_start = pynini.accep(https + colon + double_slash) | pynini.accep(http + colon + double_slash)
         protocol_end = (
-            pynini.accep("www.")
+            pynini.accep(www + period)
             if deterministic
             else (
-                pynini.accep("www.")
-                | pynini.cross("www.", "vé vé vé.")
-                | pynini.cross("www.", "dupla vé dupla vé dupla vé.")
-                | pynini.cross("www.", "kettős vé kettős vé kettős vé.")
+                pynini.accep(www + period)
+                | pynini.cross(www + period, "vé vé vé.")
+                | pynini.cross(www + period, "dupla vé dupla vé dupla vé.")
+                | pynini.cross(www + period, "kettős vé kettős vé kettős vé.")
             )
         )
         protocol = protocol_start | protocol_end | (protocol_start + protocol_end)
-        protocol = pynutil.insert("protocol: \"") + protocol + pynutil.insert("\"")
-        graph |= protocol + insert_space + (domain_graph | domain_common_graph)
+        protocol = (
+            pynutil.insert(protocol_string + colon + NEMO_SPACE + double_quotes)
+            + protocol
+            + pynutil.insert(double_quotes)
+        )
+        url = protocol + pynutil.insert(NEMO_SPACE) + (domain_graph)
+
+        graph = url | domain_graph | email | tag
         self.graph = graph
 
         final_graph = self.add_tokens(self.graph + pynutil.insert(" preserve_order: true"))
diff --git a/nemo_text_processing/text_normalization/hu/taggers/tokenize_and_classify.py b/nemo_text_processing/text_normalization/hu/taggers/tokenize_and_classify.py
@@ -69,7 +69,7 @@ def __init__(
             os.makedirs(cache_dir, exist_ok=True)
             whitelist_file = os.path.basename(whitelist) if whitelist else ""
             far_file = os.path.join(
-                cache_dir, f"_{input_case}_hu_tn_{deterministic}_deterministic{whitelist_file}.far"
+                cache_dir, f"_{input_case}_hu_tn_{deterministic}_deterministic{whitelist_file}.far",
             )
         if not overwrite_cache and far_file and os.path.exists(far_file):
             self.fst = pynini.Far(far_file, mode="r")["tokenize_and_classify"]
@@ -86,10 +86,10 @@ def __init__(
             self.decimal = DecimalFst(cardinal=self.cardinal, deterministic=deterministic)
             decimal_graph = self.decimal.fst
 
-            self.fraction = FractionFst(cardinal=self.cardinal, ordinal=self.ordinal, deterministic=deterministic)
+            self.fraction = FractionFst(cardinal=self.cardinal, ordinal=self.ordinal, deterministic=deterministic,)
             fraction_graph = self.fraction.fst
             self.measure = MeasureFst(
-                cardinal=self.cardinal, decimal=self.decimal, fraction=self.fraction, deterministic=deterministic
+                cardinal=self.cardinal, decimal=self.decimal, fraction=self.fraction, deterministic=deterministic,
             )
             measure_graph = self.measure.fst
             self.date = DateFst(cardinal=self.cardinal, deterministic=deterministic)
@@ -101,7 +101,7 @@ def __init__(
             telephone_graph = self.telephone.fst
             self.electronic = ElectronicFst(deterministic=deterministic)
             electronic_graph = self.electronic.fst
-            self.money = MoneyFst(cardinal=self.cardinal, decimal=self.decimal, deterministic=deterministic)
+            self.money = MoneyFst(cardinal=self.cardinal, decimal=self.decimal, deterministic=deterministic,)
             money_graph = self.money.fst
             self.whitelist = WhiteListFst(input_case=input_case, deterministic=deterministic, input_file=whitelist)
             whitelist_graph = self.whitelist.fst
@@ -118,7 +118,7 @@ def __init__(
                 | pynutil.add_weight(decimal_graph, 1.1)
                 | pynutil.add_weight(money_graph, 1.1)
                 | pynutil.add_weight(telephone_graph, 1.1)
-                | pynutil.add_weight(electronic_graph, 1.1)
+                | pynutil.add_weight(electronic_graph, 1.11)
                 | pynutil.add_weight(word_graph, 200)
             )
             punct = pynutil.insert("tokens { ") + pynutil.add_weight(punct_graph, weight=2.1) + pynutil.insert(" }")
diff --git a/nemo_text_processing/text_normalization/hu/verbalizers/electronic.py b/nemo_text_processing/text_normalization/hu/verbalizers/electronic.py
@@ -17,9 +17,20 @@
 from nemo_text_processing.text_normalization.en.graph_utils import (
     NEMO_NOT_QUOTE,
     NEMO_SIGMA,
+    NEMO_SPACE,
     GraphFst,
+    at,
+    colon,
     delete_preserve_order,
-    insert_space,
+    domain_string,
+    double_quotes,
+    double_slash,
+    http,
+    https,
+    period,
+    protocol_string,
+    username_string,
+    www,
 )
 from nemo_text_processing.text_normalization.hu.utils import get_abs_path
 
@@ -30,6 +41,12 @@
 server_common = pynini.string_file(get_abs_path("data/electronic/server_name.tsv"))
 domain_common = pynini.string_file(get_abs_path("data/electronic/domain.tsv"))
 
+accept_space = pynini.accep(NEMO_SPACE)
+delete_username = pynutil.delete(username_string + colon + NEMO_SPACE + double_quotes)
+delete_double_quotes = pynutil.delete(double_quotes)
+delete_domain = pynutil.delete(domain_string + colon + NEMO_SPACE + double_quotes)
+delete_protocol = pynutil.delete(protocol_string + colon + NEMO_SPACE + double_quotes)
+
 
 class ElectronicFst(GraphFst):
     """
@@ -49,36 +66,41 @@ def __init__(self, deterministic: bool = True):
         graph_digit = digit_no_zero | zero
 
         def add_space_after_char():
-            return pynini.closure(NEMO_NOT_QUOTE - pynini.accep(" ") + insert_space) + (
-                NEMO_NOT_QUOTE - pynini.accep(" ")
+            return pynini.closure(NEMO_NOT_QUOTE - accept_space + pynutil.insert(NEMO_SPACE)) + (
+                NEMO_NOT_QUOTE - accept_space
             )
 
+        hungarian_at = [
+            "kukacjel ",
+            "csiga ",
+            "ormány ",
+            "farkas á ",
+            "bejgli ",
+            "at-jel ",
+        ]
         at_sign = pynutil.insert("kukac ")
         if not deterministic:
-            at_sign |= pynutil.insert("kukacjel ")
-            at_sign |= pynutil.insert("csiga ")
-            at_sign |= pynutil.insert("ormány ")
-            at_sign |= pynutil.insert("farkas á ")
-            at_sign |= pynutil.insert("bejgli ")
-            at_sign |= pynutil.insert("at-jel ")
+            for sign in hungarian_at:
+                at_sign |= pynutil.insert(sign)
 
         verbalize_characters = pynini.cdrewrite(graph_symbols | graph_digit, "", "", NEMO_SIGMA)
 
-        user_name = pynutil.delete("username: \"") + add_space_after_char() + pynutil.delete("\"")
+        user_name = delete_username + add_space_after_char() + delete_double_quotes
         user_name @= verbalize_characters
 
         convert_defaults = pynutil.add_weight(NEMO_NOT_QUOTE, weight=0.0001) | domain_common | server_common
-        domain = convert_defaults + pynini.closure(insert_space + convert_defaults)
+        domain = convert_defaults + pynini.closure(pynutil.insert(NEMO_SPACE) + convert_defaults)
         domain @= verbalize_characters
 
-        domain = pynutil.delete("domain: \"") + domain + pynutil.delete("\"")
+        domain = delete_domain + domain + delete_double_quotes
         protocol = (
-            pynutil.delete("protocol: \"")
+            delete_protocol
             + add_space_after_char() @ pynini.cdrewrite(graph_symbols, "", "", NEMO_SIGMA)
-            + pynutil.delete("\"")
+            + delete_double_quotes
         )
-        self.graph = (pynini.closure(protocol + pynini.accep(" "), 0, 1) + domain) | (
-            user_name + pynini.accep(" ") + at_sign + domain
+
+        self.graph = (pynini.closure(protocol + NEMO_SPACE, 0, 1) + domain) | (
+            user_name + NEMO_SPACE + at_sign + domain | (at_sign + user_name)
         )
         delete_tokens = self.delete_tokens(self.graph + delete_preserve_order)
         self.fst = delete_tokens.optimize()
diff --git a/tests/nemo_text_processing/hu/data_text_normalization/test_cases_electronic.txt b/tests/nemo_text_processing/hu/data_text_normalization/test_cases_electronic.txt
@@ -1,2 +1,8 @@
 abc@def.hu~a b c kukac d e f pont hu
 https://www.nvidia.com~h t t p s kettőspont perjel perjel w w w pont nvidia pont com
+https://www.nvidia.com.~h t t p s kettőspont perjel perjel w w w pont nvidia pont com .
+@jensen~kukac j e n s e n
+www.nasa.gov.~w w w pont n a s a pont gov .
+www.enveedya.com.hu.~w w w pont e n v e e d y a pont com pont hu .
+@jensen.me~kukac j e n s e n pont m e
+@wezyr1986~kukac w e z y r egy kilenc nyolc hat
diff --git a/tests/nemo_text_processing/hu/test_sparrowhawk_normalization.sh b/tests/nemo_text_processing/hu/test_sparrowhawk_normalization.sh
@@ -1,7 +1,7 @@
 #! /bin/sh
 
 GRAMMARS_DIR=${1:-"/workspace/sparrowhawk/documentation/grammars"}
-PROJECT_DIR=${2:-"/workspace/tests/en"}
+PROJECT_DIR=${2:-"/workspace/tests/"}
 
 runtest () {
   input=$1
