Conditional Probability and Jewish Names

Also, Yahoo! Woe!

Shadal has several great comments on parashat Va’era. Let’s start the first verse:

וַיְדַבֵּ֥ר אֱלֹהִ֖ים אֶל־מֹשֶׁ֑ה וַיֹּ֥אמֶר אֵלָ֖יו אֲנִ֥י יְהֹוָֽה׃

Shadal comments:

אני ה': פועל הטובות והרעות, ורוצה אני שידעו ישראל כי גם הרע גם הטוב מידי הוא.

That is, the meaning of the name YKVK is He who performs the good and the bad. He elaborates on this point in Bereshit 2:4. To excerpt:

נ"ל כי עיקר הוראת השם הזה הוא פועל הטובות והרעות, כלומר שהכול בא מידו. גם הטוב גם הרע, וזה כולל ענין היחוד, שאין זולתו לא סיבת הטוב ולא סיבת הרע; ולפי זה לא יפול השם הזה על שום אלוה אשר איננו יחידי. ואמנם מנין לשם הזה ההוראה הזאת? נ"ל שהוא מורכב משני שמות, יה שענינו פועל הטובות, כי לא מצאנוהו בשום מקום לענין הרעות, ומה שכתוב יסור יסרני יה (תהלים קי"ח י"ח) אשרי הגבר אשר תיסרנו יה (שם צ"ד י"ב) גם שם הכוונה שהיסורין לטובת האדם, לא לרעתו; ולדעתי מלת יה או יהו מתחילת המילת הקריאה לשמחה והשתוממות על טובה גדולה, ומילת וה או הווה בהפך, קריאה וצעקה לצרה ושוד ושבר (כמו ווי vae), ושם בן ארבע המורכב משתי המלות האלה מורה פועל הטובות והרעות. וע' מה שכתבתי בשמות ו' ב' ג'.

The Yud Kay he simply says we never see associated with bad occurrences (and explains away instances in Tehillim as that the yisurim are for the person’s ultimate benefit). And this Yah! / Yeah! or Yahoo! is a an expression of joy and reaction to great occurrences. Meanwhile, Woe! of Vay! is a reaction to a bad occurrences.

He has to explain away Ehkeh, relating to being, as entirely unrelated.

וַיֹּ֤אמֶר אֱלֹהִים֙ אֶל־מֹשֶׁ֔ה אֶֽהְיֶ֖ה אֲשֶׁ֣ר אֶֽהְיֶ֑ה וַיֹּ֗אמֶר כֹּ֤ה תֹאמַר֙ לִבְנֵ֣י יִשְׂרָאֵ֔ל אֶֽהְיֶ֖ה שְׁלָחַ֥נִי אֲלֵיכֶֽם׃

This is exceptionally creative. And we can understand connections to being like Hashem by eating from the tree of knowledge of good and evil. And in context, taking this as a reaction to לָמָ֤ה הֲרֵעֹ֙תָה֙ לָעָ֣ם הַזֶּ֔ה and Hashem’s response makes some good sense. Ultimately, I’m not persuaded.

---

Shadal also discusses the names Shifra and Puah in Shemot, as the names of the midwives. In discussing whether the meyaldot ha’ivriyot were midwives for the Hebrew women vs. midwives who were Hebrew women (who naturally also catered to Hebrew women), he writes:

גם מה שטענו משמות שפרה ופועה אינם מצריים אלא משרשים עבריים, איננה טענה, כי לדעתי המילדות האלה לא היו מצריות ממש, אלא מארץ גושן שהיתה קרובה לארץ כנען (עיין בראשית מ"ו ל"ד), וקרוב שהיו מדברות שפת כנען ושמותן כנעניים ועבריים.

That is, in response to those who point to the Semitic rather than Egyptian character of the name, he dismisses the argument. For these midwives were not actual literal Egyptians, but were (non-Jewish, Egyptian-affiliated) from the land of Goshen, which was close to the land of Canaan (see Bereishit 46:34), so it is likely that they spoke Canaanite, and their names were Canaanite or Hebrew names.

To expand on this point a bit more, see W. F. Albright’s article, Northwest-Semitic Names in a List of Egyptian Slaves from the Eighteenth Century B. C. They discovered Egyptian slave lists “dated in the first and second regnal years of Sekhemre' Sewadjtawi Sebekhatpe III, about 1740 B. C. There are 95 slave names, 37 of which are labeled as Semitic (though a few of these are really Egyptian, as we shall see).”

These names include #13, Sk-ra-tw, which is parallel to Yissachar.

and #11, Mnhm, an abbreviated form of “(Such and Such a God) Shows Mercy”, parallel to Menachem.

But the one of interest to us is #21, the parallel to Shifra.

Albright writes:

By hypocoristicon he means shortened name. Note, Puah does not appear on the slave list. Later in the article he writes:

3. The name of one of the Hebrew midwives, "Shiphrah" (Ex. 1: 15) appears here in nearly the same form (No. 21), and the other name has already been duplicated through Ugaritic finds, though not hitherto recognized. “Shiphrah" and “Puah" are thus perfectly good Northwest-Semitic names of women from the first half of the second millennium, and there is no reason to consider them as not authentic in their appearance about the end of the 14th century B. C. in Hebrew tradition.

---

In Va’era, we encounter the name Putiel:

וְאֶלְעָזָ֨ר בֶּֽן־אַהֲרֹ֜ן לָקַֽח־ל֨וֹ מִבְּנ֤וֹת פּֽוּטִיאֵל֙ ל֣וֹ לְאִשָּׁ֔ה וַתֵּ֥לֶד ל֖וֹ אֶת־פִּֽינְחָ֑ס אֵ֗לֶּה רָאשֵׁ֛י אֲב֥וֹת הַלְוִיִּ֖ם לְמִשְׁפְּחֹתָֽם׃

Shadal writes:

פוטיאל: נראה מורכב ממלה מצרית פוטי (עיין בראשית ל"ז ל"ו) ומשם אֵל בלשון הקדש, וענינו מיוחד לכבוד האל (בני בכורי זצ"ל), וגיז' פירש Afflictus a Deo.

That is, it is a hybrid term, with the Puti being Egyptian and the El being Hebrew. (He credits his firstborn son, Ohev Ger, Filosseno for this explanation.) Bereshit 37:36 is where he explains the name Potifar:

פוטיפר: בתרגום יוני Petefre, ואומרים האחרונים שענינו הלשון מצרי מיוחד לשמש (qui solis est, idest soli proprius et quasi addictus) וכן Petamun מי שהוא מיוחד לאמון (שם אליל), Patusiri מי שהוא לאוזירי.

---

I had an intuition, which is now perhaps wrong. That intuition is that just as Poti-far and Puti-el are Egyptian names, we could guess or calculate whether a name is Egyptian or Hebrew based on Unigram or Bigram probabilities, forgetting about etymologies. And further, that Shifra and Pu’ah would then be good Egyptian names.

This is a classic NLP task. For instance, in the NLTK book, they classify names as male or female based on text features. Here is their simple example, from Chapter 6, Classifying Text. We write a feature extractor function, which will take in a word (a name) and return a dictionary where the key “last_letter” is associated with the actual last letter of the word.

We then read in 500 male names and 500 female names, such that we know the correct labels. And we reduce the full text of the names to their feature set dictionaries.

Next, we train a Naive Bayes classifier that will predict the most likely tag given the feature:

Finally, we can use that classifier to predict the gender of people based on their names. For instance, Neo and Trinity from the Matrix.

This classifier has about 77% accuracy on the test set. And it uses only a single feature, the last letter of the name. If you think about it, this makes some sense. Think of female names such as Sara or Jessica, Cindy or Tiffany. Then think of Eric and Bob. Compare Robert and Roberta. The last letter can be highly indicative.

As we add more features, we can obtain greater accuracy. The features I would like to use for Egyptian and Hebrew are just the presence or absence of specific letters, and the presence or absence of specific runs of letters.

Think about Jewish names you know. Moshe. (Or is this Egyptian?) Shmuel. Baruch. Yaakov. Think of secular names you know. Philip. Robert. I am excluding Biblically based names, but if you want, think of Samuel, Moses, Benedict, Jacob. Specific letters or letter pairs appear or don’t frequently appear. (J in Hebrew names, CH in English names.) So we built a classifier based on these features, and see if a name sounds Jewish.

Wikipedia has a list of ancient Egyptian names, and we can obtain a bunch of Biblical names from genealogical lists. Then, we can predict whether Shifra and Puah are Egyptian or Hebrew.

I would have guessed that the individual letters Peh, Tet, Ayin and Resh are more likely in Egyptian. Think of Pharaoh, Potifar, Potifera, Tzafnat Paneach, Pash-chur, Harnepher. Think Achi-Ra’ and the name Ra’. So Shifra had a bigram of PR, and Puah had a P-’ . But there seems to be evidence contradicting my assumption.

If I get around to implementing this, I will update.