Featured Linguist: Stefan Müller

I am delighted to support this year’s fund drive for the LINGUIST List.

While preparing this text, I had a look at the pieces from previously featured linguists and noticed that I share some characteristics with Adele Goldberg and Colin Phillips: we all had a passion for mathematics. I was a member of the Mathematische Schülergesellschaft (MSG) run by researchers from the Humboldt University from the fifth grade onwards. When I was 13, I applied to the Heinrich Hertz Oberschule, which is a school with a specialization in mathematics (nine hours of math each week). Back then two to four pupils out of 30 could go to the Extended Secondary School and getting a place on this special school was even more competitive. There were two tests: a math examination, which I finished with 100%, and a political talk, which I failed. They asked me whether I would want to serve in the army for an extended period of time (three years instead of one and a half), and I told them that I never thought about this question but that I thought it was a bad idea. Since the GDR expected loyalty of those who were allowed to these extended schools and those who were allowed to study, I was rejected. I am very grateful to my parents who left no stone unturned in order to get me into this school. They got certificates from my math teacher and from the MSG, and I had a second chance interview on political issues with the principal of my school. I told them that it was my deepest wish to serve in the army for three years (sarcasm).

My application having been successful, I went to the Heinrich Hertz school, and it was great. Lots of math and even computer science. We could learn to program using programmable pocket calculators from Texas Instruments (note that GDR money could not buy you those, so the school must have had some special connections), and later using the first home computers produced in the GDR (KC85-1). The school organized a partnership for individual pupils with the Humboldt University. I could work in the main computing facility of the HU. I was very privileged: while university students had to use punch cards at the time (1985), I could type in my programs at the terminal of the HU’s main computer. This mainframe had 128k of memory and an electronic typewriter typing out important messages. Clack, clack, clack. (/dev/console)

After this happy childhood with mathematics and mainframes, I had to serve in the army. I hated every minute of it; it was the darkest period of my life. I went to the library in the town where I was stationed and got a book by Kurt Schwitters: “Anna Blume und andere”. Kurt Schwitters is one of the proponents of Dada (an art movement of the European avant-garde in the early 20th century) and this was the right kind of craziness for me. I knew immediately: either I write crazy stuff like this or I will really go crazy. So I went for Dada. I published my stuff together with a friend who did illustrations in a Samizdat, an underground self-publication, using army computers and printers. I wrote the publishing and layout software myself, a precursor to what I would do later in life.

Finally released from the army, I began to study at the Humboldt University. I studied Mathematical Computer Science, a brand-new subject, which was a full-scale math curriculum with computer science on top. I started in September 1989, when GDR still existed. The elections on March 7th of that year were fake, and people left the country in masses via Hungary and Czechoslovakia. The situation culminated on October the 7th, which was the 40th anniversary of the foundation of GDR but also one of the days on which the opposition kept reminding the government of the fake elections which had taken place on another 7th day of a month. Protesters and celebrators mixed, and riot police with helmets and water guns were in the streets: something that had never happened before. Secret service members mixed with the crowd and tried to stop discussions, arresting many of the protesters. The outcome of all this is known. There was a huge demonstration in November and things began to change slowly. A world collapsed and something new began. I was deeply disappointed by the faculty members: they just continued to talk about algebra, analysis, and logic as if nothing was happening. They completely ignored the outside world. Their math was dead. It worked as it always did, it was correct, but it was boring. I could not connect to them any longer. I wanted something different, something dirty, something that is not perfect: language. One thing the Dadaists did in the 1920s was anagrams. They made anagrams with pencil, paper, and scissors. Being a nerd, I went for more systematic and efficient methods: I used a computer. We learned Prolog in the computer science lectures, and so I used Prolog to permute the letters, and a dictionary to then test the permutations against word sequences. Of course arbitrary word sequences were not good enough; I wanted to filter out all those sequences that are naturally occurring phrases. So I wrote my first grammar of German.

Udo Kruschwitz, a friend of mine, had the idea to go to Great Britain for a year. We checked Glasgow, Edinburgh, and some other places, and decided to apply for Edinburgh since they did not require a language test. My English was terrible at the time, but as far as computer science lectures were concerned, I did not have any problems. As so often in my life, I was lucky: Edinburgh was a hot spot of computational linguistics. Chris Mellish, the pope of parsing and Prolog, was there. Alan Black did computational semantics. Henry Thomson did formal linguistic stuff. It was just great. It was an international mix of students, including David Adger, and it was really inspiring with a lot of fruitful interaction. The year abroad ended with a Large Practical in which we used Prolog for syntactic and semantic analysis of English. I did a PATR-like grammar with a Discourse Representation Semantics component.

When I came back, the courses I did in Edinburgh were accepted so that I could finish my study in four years including the year abroad. The original plan for my diploma was to develop a grammar and a processing system that incorporated the ideas about verb fields by Jürgen Kunze (Professor for Computational Linguistics at the Humboldt University back then). This theory is a very cool theory since it uses semantic primitives like ’cause’, ‘become’, and ‘have’ to model the verbs of exchange of possession (‘give’, ‘take’, ‘steal’, ‘lend’, ‘borrow’). Kunze modeled the field in German: 91 verbs. A 92nd verb would have been expected, so he discovered a lexical gap in German.

But, back in Edinburgh, I had visited one of my supervisors. There was a copy of the first volume of Pollard & Sag about Head-Driven Phrase Structure Grammar (HPSG) on his desk. I asked him what this was about, and he said: “Oh, this is too difficult for you.” I knew how to work with Definite Clause Grammars, but I also felt that a bit more than this would be required for my diploma thesis. So I had a look into Generalized Phrase Structure Grammar (GPSG) and developed a parser for a fragment of German. I compiled out all phrase structure rules, and the result got so messy and dealing with the grammar got so complicated that I decided that HPSG could hardly be more complicated than what I was already doing. So I turned to the forbidden book, and I loved it. I also got an ESSLLI booklet from 1992 with a draft version of the ’94 HPSG book. I read the books several times. I developed an HPSG grammar of German and a parser for it. Sometime in December 1993, Kunze said that there was a position I could take. He asked what I had done so far for my diploma and decided that this was enough. So I never got to his cool semantics but was deeply involved with HPSG.

Kunze was coming from the Academy of Science of the GDR, which was closed down after the end of the GDR. A lot of scientists from GDR universities or other research institutions became unemployed (half of the 218,000 researchers in GDR, two thirds of the professors). But there was also a program called Wissenschaftlerintegrationsprogramm (researcher integration program). The idea was to give three-year contracts to East Germans so that they could find their place in the West German academic system (or rather in the German academic system). There was money in the pool for Kunze’s position, but since he got a professorship, this money was not needed. He used it, together with other money, to create a position, which I was lucky to get. This was a huge privilege, since it was a personal position with a personal travel budget of 10,000 Deutschmarks, and I was freshly graduated, not taken over from an existing GDR institution. I used the time well to extend the HPSG Grammar and improve the parser. By the end of the three-year period, I had the largest grammar of German available and the fastest system for processing it. As Wolfgang Wahlster, one reviewer of my PhD thesis, remarked: It was better than what IBM came up with, with a research group working on grammar and parser and a 1 Mio DM budget. But when I presented my work to the computer scientists of the Humboldt University, they did not really like it. I guess I made a mistake in presenting the stuff; they were not interested in the grammar part at all. So, rather than trying to convince them that computational linguistics is interesting and that I had something that is worth a PhD, I looked for different options, and so Hans Uszkoreit and Wolfgang Wahlster became my PhD reviewers. I defended the PhD at the computer science department in Saarbrücken. I also got a job there at the DFKI (German Research Center for Artificial Intelligence) in the Computational Linguistics (CL) lab of Hans Uszkoreit, working in the VerbMobil project. The goal of VerbMobil was speaker-independent machine translation of spoken language. Nowadays every phone can do this, but CL was far away from this goal in 1992, when VerbMobil started. The project was way too big, as far as the number of involved people was concerned. Wolfgang Whalster counted 911 people (including student assistants). Many complained that communication was difficult, but for me the VerbMobil time was just great. Again, I was lucky: everybody who was directly before or after me in the processing pipeline was in my lab. There were Uli Krieger and Bernd Kiefer, responsible for parsing, and Walter Kasper, with whom I interacted closely since he did the semantics for my grammar. If something was wrong with the output format or the semantic representation, they got the message from other groups and we dealt with the problems internally.

After defending my PhD, I wanted to give a conference talk based on the chapter about particle verbs in my thesis. The reviewers did not like it. I felt hurt and misunderstood. I decided to write something bigger, maybe a journal article. Particle verbs are complex predicates. They form a predicate complex similar to verbal complexes and resultative constructions with adjectives. In order to write this up properly, I had to talk about secondary predication (depictives and resultatives), about copula constructions, and about verbal complexes. All these phenomena interact with fronting, passivization and so on. I realized quickly that I had to write a book. This became my habilitation. Another lucky coincidence in my life.

VerbMobil ended in 2000, but this was the time of internet startups and the DFKI was involved with several of them. One was Interprice (now Semantic Edge), a platform for price comparison with a natural language interface. The domain was later changed to travel. I continued to develop the German grammar for these domains and was responsible for other languages too. In 2001 I got the offer to replace the chair for computational linguistics in Jena for two years.

When this time was over, I was lucky again and got an assistant professorship for Theoretical Linguistics/Computational Linguistics in Bremen. After a year in Potsdam, filling in a position in CL, I got my first permanent position at the Freie Universität Berlin in 2007. This position was in German and General Linguistics, and while my focus had been on German in previous years, I now started projects on Danish, Persian and Mandarin Chinese. I developed computer-processable HPSG grammars for German, Danish, Persian, Mandarin, English, Yiddish, Spanish, French, Maltese and Hindi. All grammars come with syntax and semantics, and some with an information structure component which was developed in the Collaborative Research Center 632. The grammars share a common core grammar, that is, constraints that hold for all (examined) languages or for subgroups of them. In 2016, I moved to the Humboldt-Universität zu Berlin, where I have had a chair in German Syntax since then. We started another Collaborative Research Center there (CRC 1412). It deals with register phenomena. My lab contributes a project dealing with the empirical side (corpus linguistics) and with questions on how to pair HPSG with probabilistic aspects.

Throughout my life I spent a lot of time comparing various theoretical frameworks. Most energy went into the lexical vs. phrasal constructions discussion (basically phrasal Construction Grammar vs. HPSG/SBCG/Categorial Grammar/Minimalism) and I really enjoyed working together with Steve Wechsler on a target article, but there are also other comparisons like the ones of HPSG and Minimalism as well as HPSG and Dependency Grammar. In general, I think that linguistic frameworks share a lot of ideas, and we should talk more to each other to be able to understand the commonalities and work together across framework boundaries.

The Linguist List has been important to me throughout my scientific life. It played a role in scientific exchange (I remember the challenge to the Minimalist community to build a parser showing that Minimalist ideas can be made consistent and working.


The resulting discussion was very interesting…), the job postings were very important, and one early post by Martin Haspelmath (in 2004!!!) to the list about open access turned out to be one cornerstone in the foundation of Language Science Press.


During a dinner in 2012 with Adele Goldberg, Thomas Herbst and Anatol Stefanowitsch, we realized that the tools for running a scholarly-owned open access publishing house were now available. I started writing to scholars asking for signing at a website for support. I remembered the post from Martin and asked him if he wanted to join the enterprise. We met at the Freie Universität Berlin, talked things through, wrote a DFG proposal, founded Language Science Press, and the rest is history. (Part of this history, and an important one, is Sebastian Nordhoff, who worked for Martin and later joined Language Science Press. Without him and his social and technical skills, the great success of this initiative would not have been possible.)

The organizers of this year’s fund drive of the Linguist List gave us a motto for our piece: silver lining. We all need one, after over a year of Covid, which has been exhausting and terrible for everybody in academia (and beyond). And this is not the only crisis we are in. The climate crisis is something that is not waiting for us at the end of the pandemic. It is something that is happening simultaneously and on top of everything we already have. I have been active with Scientists 4 Future for several years now. Martina Schäfer, Gisbert Fanselow and I started an initiative for flying less, and we collected signatures of people who pledge to refrain from business flights to destinations less than 1000km away. 25% of the scientific staff of the Humboldt University signed. Given the fact that 50% of the carbon emissions of universities is due to travel, and 93% of this is due to air travel, this was a good result. However, we are far from zero carbon emissions, and we have to go down to zero. There is not much time left (as the most recent report of the IPCC made clear again).

So, where is the silver lining? Well, maybe it is not easy to see it in all the smoke coming from the fires in Canada, Spain, Italy, Greece, and Turkey. But once the smoke clears, we may be able to see that Covid brought us one thing: we learned to live without traveling. Search committees, job talks, conferences are possible without intercontinental flights. The tools we are using are not perfect, and I experienced disasters in communication which I was very unhappy about, but we can improve on this and learn how the online world works. Since we must go down to zero emissions rather quickly, it was good that we’ve now learned, at least in principle, how to do it.


Leave a Reply