Friday, March 01, 2019

Early Career Academics: Tips for Preparing for your Interview Visit

This post goes out to early career academics about to interview for positions. Preparing for your interview is very different than preparing for just another seminar visit. Moreover, the observable difference between a well-prepared candidate and one who sees the interview as  just another invited seminar visit is HUGE. Do yourself a favor and ask faculty you know how to prepare. Most faculty, even junior faculty, have served on at least one search committee and will have a lot of good perspectives from sitting on the other side of the interview process.

This post gives a few tips that I've picked up after sitting on search committees. Your mileage may vary, and you should definitely ask around. But this is what I notice...



Given that your audience will be with everyone you meet that day as well as everyone who sees your presentation (which may be video taped and viewed in a tiny screen, which is something you might want to consider when crafting it), you should survey several people to see what sticks out for them about good and bad candidates. Ultimately, people make up their minds very quickly based on relatively small things. The search committee might have combed through your CV, and certainly some who submit feedback to the search committee about your visit will have as well, but many will make their decision during your talk (or multiple talks, in the case of some departments) and/or in the few minutes they have had to talk to you during a half-hour meeting or over a meal. So get a good idea of what "many people" might like.

So below I've put some good general tips from my perspective. Again, your mileage may vary.

FIRST, THE NON-TALK-RELATED TIPS (talk tips in next section)

It's not all about the job talk. I've given job talk tips below, and the job talk is extremely important -- one of the most important talks you'll give. However, there is a lot of the day(s) that won't be the job talk, and you need to be prepared for all of the interactions you'll have. So I put the non-job talk tips first.
  • DO YOUR (VIRTUAL) RECONNAISSANCE IN ADVANCE! When you find out you are going to be interviewing, you may not know who exactly you will meet that day. Even when you're given the schedule, there may be last-minute additions. So if you can, do your best to memorize who all of the faculty are and know roughly what they do (up to a 2-sentence description). You may not know who the search committee is, but you should use whatever clues have been made available in the communication thus far to be able to spot them in a crowd so that you know they're coming. You never want to be in the position where you have to ask someone what they do. You want to make it seem like you are genuinely interested in their research and can ask intelligent questions about it. You also want to try to anticipate what they find valuable and adjust your answers to questions to be complementary to these perspectives.
  • Try to understand how the school is structured and where your position would fit in that structure. Try to understand what resources are available. Be able to tell someone what specific things attract you to that school or that program. Make it look like you've been waiting for this particular position to open and jumped on the opportunity as opposed to just interviewing at any random school.
  • Be very courteous and approachable and never dismissive. Be comfortable. Do your best to be relaxed and conversational while also internally keeping up a little bit of a guard. You don't want to be too quiet, but you want to be more professional than you usually are on an average visit. You aren't required to answer certain classes of personal questions (whether you have a spouse, kids, plans for these things, etc.), and so you can divert if they happen to come up casually, or you can try to use them if you think they would be an advantage. But don't give up too much information about things that could make it difficult for you to accept a position. Those tricky details can come out during negotiation.
  • Keep in mind that you want faculty to want to work with you. You want them to be excited about collaborating with you.
  • Keep in mind that you will also be evaluated based on whether there is too much overlap with existing faculty. Do your best to emphasize the new things you can bring (without accidentally pointing out weaknesses in other faculty or the program as a whole).
  • You may be given 30 minutes alone with graduate students. Prepare for this time by having a set of questions that you can ask them (in case they don't have much to ask you). You have already studied up on all of their advisers, and so you should be prepared for the different research directions they have and can maybe anticipate some of their answers. But let them speak. You want them to see you as another faculty member that they would want around as a resource. Their advisers may ask them how they felt about the interaction.

JOB TALK TIPS:

And here are the general comments about job talks. See the previous section about non-job-talk related tips (in short: DO YOUR RECONNAISSANCE and BE COURTEOUS)...
  • Practice your job talk before hand so that your timing is flawless and you are clearly confident with the material. And get your timing and talk density right so that you END WITH TIME FOR QUESTIONS. Ten minutes is sufficient. But don't leave more than 15 minutes. But if you leave less than ten minutes, people will not be impressed (and may even be annoyed).
  • Make sure your job talk tells a consistent narrative that gives everyone in the audience an idea of what your research vision is and how your career up to this point has successfully implemented that vision. This may not be true a priori, but you need to find a way to tell your story to make it seem true. People don't want to see a random collection of research. They want to use the talk to get to know you, what you've done, what you will likely do in the next 3–5 years, and be impressed with both the current body of work and the potential. With that in mind, you don't have to present everything if some things don't contribute to the broad overall narrative. If you still have some work you're proud of that you don't think fits a narrative, you can include it briefly as a kind of "Other things that I am interested in" near the end to show you have breadth. But don't hop back and forth between disconnected projects. People will forget what you're all about and get frustrated by the lack of consistency.
  • If your work has been published or presented at major conferences, call out these venues in your presentation as you go through them. You want people to be convinced that other people care about your work. Don't hit them over the head with giant bibs, but maybe include a small parenthetical ref at the bottom of slides here and there and then say things like, "In work I presented at .... last year," or, "In work that came out in ... a few months ago..."
  • If you have more work than just your PhD work, be sure to show it somewhere. People like to get a sample of what you'll do as an independent researcher. Sometimes just the PhD work doesn't quite capture that.
  • Include at least a few slides on future directions. You don't want people wondering what you'll do next. Pointing out where you have already received funding is a good thing, and it's definitely good to identify where you'll go next for funding (agencies or even particular programs). Some sort of flowchart showing how your research vision can be divided into specific threads that meet the objectives of these different agencies is great as it is more convincing that your vision can be operationalized for the next 3–5 years.
  • Figure out if you are in a discipline (or even a school) that cares so much about education that a significant portion of your job talk should be associated with your classroom innovations and perspectives.
  • When you get to the end of your talk and are taking questions, try to maximize for the quantity of questions. Don't dwell on one particular answer, and don't give one question too much time. Do your best to respectfully acknowledge the value of the question you were given, but try to table long discussions for "off line" discussion. This is the only time some faculty will have to interact with you, and you don't want to frustrate some who have questions by being too thorough with someone else's question. It also looks much better if you answer 3 questions than spending a long time on 1. So find a way to pivot quickly to another question if your first question is starting to take too long and prevent you from getting to others.
  • The Dean/Head of School/Director may be in the audience. They may have a question. Take that question first as they may have to leave earlier than everyone else, and they also have an outsized role in hiring decisions.
Make sure you go to a bunch of job talks before your own job talk to see the diversity and try to guess which ones are good and bad job talks. Usually (but not always) more senior people give better job talks because they already have a good idea of what is good and bad because they've done it more and been a part of the decision-making process themselves. Contrast these more senior researchers with junior researchers who are definitely giving their first job talk. You'll notice consistent differences in the structure of the talk. I would say a good structure is something like...
  • Here's my vision
  • Here's a few projects that fit well together that show that vision
  • Here is where that vision takes me in the future and who will pay for it
  • Oh, by the way, here are other things I do too, but I don't have time to go into in detail
  • Here are my perspectives on education (depending on the school/discipline, you may not have this section at all or it may be half of your talk)
  • Here are 10–15 minutes I've made available for questions

Tuesday, October 10, 2017

Review: Don't buy the JBL Link 10 or the JBL Link 20! Junk speakers have unavoidable ugly popping noises.

TL;DR

The JBL LINK 10 and LINK 20 should have been amazing products. Unfortunately, all three units purchased from two different Best Buy stores have the same major audio problem. The audio cracks and pops even at medium volume of the most delicate content. I cannot get through Clair de Lune (at medium volume) or even streaming of voice content from NPR without very noticeable popping at random instants – the kind of popping you might expect if you had a loose connection within a speaker amplifier that generates discontinuous movements of a speaker every time it vibrates loose for a moment. We have returned all three speakers. I do not recommend you waste your time or money with these things.

I was originally very excited by the announcement of the JBL LINK Series, with the JBL LINK 10 (below, left) and the JBL LINK 20 (below, right) models being particularly interesting to me.

JBL LINK 10: New Google-Assistant-enabled speaker from JBL (available at Best Buy)            JBL LINK 20: New Google-Assistant-enabled speaker from JBL (available at Best Buy)

In short, both speakers are like portable versions of the traditional Google Home with the promise of potentially better audio quality because they are made by speaker companies. There have been several 3rd-party Bluetooth speakers with Amazon Alexa integration, but they are more similar to the Echo Tap than the full Echo because they required pressing a button to call up Alexa. Instead, members of the JBL LINK series are fully voice activated just like Google Home, and the LINK 10 promises a 5-hour battery life when unplugged and the LINK 20 promises a 10-hour battery life when unplugged. Furthermore, both the LINK 10 and LINK 20 have an IPX7 waterproof design which means you can not only take them outside but you can put them in the shower. My wife and I have been looking for an elegant way to extend our voice assistants and multi-room audio systems outside, and this new line of speakers seemed perfect for that. Plus, when we brought them inside, they could serve as full-featured voice-activated Google Home replacements (albeit without the beta features that we get early on our real Google Homes).

On the day the JBL LINK 20 was released, we purchased a $199 JBL LINK 20 from a local Best Buy. Setting it up was identical to setting up a Google Home. The unit came with a JBL's characteristic orange 10W USB charger and micro-USB cable; however, the unit had plenty of charge and could be used off the charger out of the box. Once I set it up, I asked Google Assistant to stream from our local NPR station. I was surprised to hear subtle but very noticeable popping at random times during the stream. I thought this could be a problem with the stream (although I did not notice these popping noises when I streamed on true Google Home or Amazon Echo devices in our home), and so I played a nice song called "Tennessee Whiskey" by Chris Stapleton, and I noticed the same problem. Finally, I played the gentle, all-piano "Clair de Lune" by Claude Debussy at medium volume (2 or 3 dots on the front of the unit) and heard the same random popping. The popping is unrelated to whether the speaker is plugged into the charger or not; it is consistent in all operating conditions.

Being an optimist, I assumed I must have received a dud. Trying to minimize the chances of getting another from a batch of duds, I returned the unit back to a different Best Buy and purchased BOTH a $149 JBL LINK 10 and a brand new $199 JBL LINK 20 to try them out. I was surprised to find out that the $50 difference between the two not only paid for the larger speaker and battery but also paid for the USB charger! Although the LINK 20 comes with a 10W charger, the LINK 10 only comes with the cable. Regardless of that, they BOTH had the exact same problem as the first LINK 20 we purchased – they couldn't pass the "Clair de Lune" test at medium volume! So I have to conclude that there is something fundamentally wrong with the JBL LINK series (at least in the LINK 10 and LINK 20 varieties).

I'm extremely disappointed by this result. It makes me think that the sale of Harman (and JBL) to Samsung earlier this year is starting to take its toll on product quality. I guess I should be happy the device didn't explode in my hands. Still, I'm sad it doesn't provide the awesome sound quality of the much older JBL products I own. Even the $80 JBL FLIP 4's that I have clearly out-perform these much more expensive LINK 10 and LINK 20's. I was thinking about pairing the new Google Home Mini with a JBL FLIP 4, but I just learned that the Google Home Mini has no 3.5mm audio jack and only has Bluetooth support as a target speaker and not as a source (so you are stuck with the Mini's apparently crappy speaker). If only the Chromecast Audio could directly connect to a Bluetooth speaker (without some ugly dongle), then at least I'd have a path toward an elegant solution to outdoor casting. Meanwhile, back to the drawing board!

Thursday, April 13, 2017

Half the reason for POMDP's in the literature: You can pronounce the name!

Today, I noticed a new paper in Mathematics of Operations Research, which suggests a way to combine risk-sensitivity with POMDP's without having to use only exponential utility functionsI'm bothering to write this post because the title is infuriating to me. In my opinion, the title of the paper will probably largely contribute to why it will be forgotten – only to be referenced by other papers that find it while doing their post hoc due diligence

"Partially Observable Risk-Sensitive Markov Decision Processes"
by Bäauerle and Rieder
Mathematics of Operations Research (2017), Articles in advance

The great success of POMDP as a framework used in the literature is in part because people like the name – "POM-D-P". If you're taking in a young graduate student, you can send them to the literature with a few quick words – "Go check out POMDP's". 

But what am I going to do with PORSMDP? "Poor-Sim-Dop"? "Po-Rism-Dip"? Even if one of those managed to roll of off the tongue, the relationship to POMDP wouldn't be obvious.

Why didn't anyone in the chain of custody of this manuscript suggest putting "Risk-Sensitive" up front, as in RSPOMDP (R-S-POM-D-P)? Or maybe at the end, as in POMDPRS ("POM-DiPpeRS")? The latter suggestion not only is memorable, but it sounds delicious.

Just some food for thought.

Tuesday, March 22, 2016

Phone scam: "Enforcement action" from "US Treasury" (302-279-3069)

At 8:00 AM this morning, I received a phone call from 302-279-3069 that I let go to voicemail. On the voicemail, there was a threatening but obviously scripted monologue along the lines of...
I'm calling regarding an enforcement action executed by the US Treasury requiring your serious attention. Ignoring this will be an intentional attempt to avoid initial appearance before a magistrate judge or Grand Jury for a federal criminal offense. My number is 302-279-3069. I repeat 302-279-3069. I advise you to cooperate with us and help us to help you. Thank you.
I especially like the "help us to help you" part.  This is basically the same phone scam reported by many others in tax seasons.

I encourage anyone with free time to give this number a call (dial *67 before you call them to mask your caller ID). If they're on the phone with you, then they won't be on the phone with some poor person that they've tricked.

Monday, January 19, 2015

A New Way to Download ePub NOOKBooks from Barnes and Noble

Back in September, Barnes and Noble removed the ability to download DRM'd ePubs directly from their site. If you are still using NOOK for PC or NOOK for Mac, you might be able to use them to download your newly purchased books and then find the ePubs in some obscure directory on your machine. However, that doesn't help Linux users. Moreover, those Desktop apps are not supported anymore, and so it is not an ideal solution.

 Barnes and Noble NOOK Books

So what do you do? If you have an Android device (in principle, you could use an Android emulator as well), you can download the NOOK e-reader Android app and, through a few steps, get the same ePubs you used to download. Here's how you do it:

  1. Download and install the NOOK e-reader Android app on your Android device
  2. Open the NOOK e-reader app and find your library with your newly purchased book
  3. Select "Download" to cache the book locally on the device, and wait for the download to finish
  4. Once the download is finished, open a file manager application like ES File Explorer and navigate to
    /sdcard/Nook/content
    You should find your NOOK book ePubs in that folder. That is, find the Nook folder in your internal storage, and then find the content folder inside it.
  5. You can use your file manager to copy those ePubs out of that folder, or you can "Share" them to another installed app like Dropbox or Google Drive so that you can easily get them on other devices or your Desktop computer
Once you have the ePubs on another device, you can do whatever you used to do with them when you could download them directly from the NOOK library on-line in your web browser.

[ Side note: You might try to plug your Android device into your computer and use MTP to access /sdcard/Nook/content directly from your Desktop and copy files out of it. However, it's possible that the ePubs won't be accessible this way. In fact, the might not even show up in the file listing. So you might be stuck using a file manager on your device. ]

Tuesday, January 13, 2015

Dollar Shave Club: My review and a cheaper alternative

I recently signed up for the Dollar Shave Club (DSC), which is a site that lets you paying $6 every 1 or 2 months for 4 razors by mail (and you can pause your subscription at any time). It turns out that you can get the exact same razors for much cheaper with no monthly subscription so long as you're willing to buy in bulk. Apparently DSC gets its razors from Dorco USA, which you can buy directly from on-line. In fact, you can buy retail handles with trial razors from Dorco too. For example, for $6, you can try out a Dorco handle and two razors, and then you can buy 36 replacement cartridges for less than $1/razor.


I tried my "4X" DSC (or Dorco USA) razor for the first time this weekend. It was an OK razor. It had a nice heavy handle, but the angle of the pivoting head was not as steep as I'd like, and the moisturizing strip was far too thick. Consequently, it was very awkward to get into small nooks -- like under the nose. The "4X" razor doesn't have a trimming edge either (the more expensive 6-blade "Executive" model does), which would really improve matters. But after the shave, my face felt pretty good. It just took a lot longer to get through a full shave. So I'm not sure I can fully recommend these razors over, say, Schick Quattro Titanium disposable razors, which actually have a trimming edge and are pretty cheap from local stores or from Amazon (around $7/3 razors).

DSC also sent me a tester of their "Dr. Carver's Easy Shave Butter," which actually burned my face a bit as it went on (but I have sensitive skin). My face was fine afterward. But using Nivea For Men Active3 Body Wash for Body, Hair & Shave worked just as well without the sting. So I'm not sure I can recommend the Shave Butter either.

So I'm torn. It's possible that the 6-blade+edging "Executive" unit will be better. Either way, in the long run, it's cheaper to buy from Dorco USA directly, and I like the idea of dropping the subscription service. Dorco USA also sells other razor models and even disposable models. They also sell razors designed for women, whereas DSC just tells women to buy the 4X. The 4X has a handle which is somewhere between what you'd want for the face and the leg, but I think it's biased toward face shaving. So the lady's models at Dorco USA are definitely recommended if you need to shave legs.

So that's my brief review. tl;dr -- Check out DORCO if you like DSC's blades but don't like the subscription service.

References:

Monday, December 29, 2014

Looking for a cheap Bluetooth-capable portable MP3 player? Buy a no-contract smartphone and don't activate it!

My father needed a portable Bluetooth-capable MP3 player with large or expandable memory. Portable MP3 players with such features typically sell for $80+ (Trio mini), with the nicest units being at least $150 (iPod nano) or much more. The cheap models have terrible battery life (3 hours for Trio mini). Creative solution? He bought a no-contract LG Optimus Exceed 2 from Verizon Wireless for less than $50 at Best Buy and used a 16GB microSD card that he already owned.
LG Optimus Exceed 2, Prepaid (no-contract) Phone from Verizon Wireless
This particular no-contract phone is fast, runs Android 4.4.2, and can be used without activating. In fact, it is very easy to bypass the initial activation screen with a special key sequence – volume up, volume down, back, home – pressed after restart at the language selection screen. After bypassing that screen just once, the phone is happy to run unactivated indefinitely after that – it will never present that activate screen again. Use airplane mode to prevent phone from looking for a network to save battery, and flip on WiFi and Bluetooth. It's like a $50 iPod touch. Nice deal.

You can get this phone from a number of places, as shown in links below. Of course, there are other no-contract phone choices you could make (check all no-contract carriers – Boost Mobile, Virgin Mobile USA, T-Mobile, etc.). If you're willing to compromise on the phone software or other capabilities, you can get decent "MP3 player phones" for $30 or even $20.
For example, you can get essentially the same phone from Boost Mobile packaged as the LG Realm for only $35 on Amazon right now. It's $40 from other retailers.
LG Realm, Prepaid (no-contract) Phone from Boost Mobile
So that's like a $35 iPod touch. Even nicer deal!

Monday, February 04, 2013

LaTeX template for NSF-style Biographical Sketch

On large multi-university NSF grant proposals, NSF requires that senior personnel submit a 2-page biographical sketch ("biosketch") that is formatting according to certain rules in their Grant Proposal Guide (GPG). The format is pretty simple, and so there does not seem to be much demand for a solid LaTeX template for one. Nevertheless, I thought some people might find one helpful. I've posted a PDF of my NSF-style biosketch below along with the TeX source used to generate it. On a related note, you can also find my curriculum vitae (CV) tailored for faculty searches (PDF, TeX) as well as for industry searches (PDF, TeX).

Thursday, January 24, 2013

Discrete-time Phase Portraits?

I was contacted recently by e-mail asking how to produce a phase portrait of a discrete-time system. In my initial response, I explained that a true "phase portrait" wasn't defined for discrete-time systems because the technical notion of a phase portrait depends on a special structure that comes along with ordinary differential equations. The original poster needed some additional clarification, and so I sent a second e-mail that I have posted below. It touches a little bit on the original poster's question, it comments on differences between discrete-time and continuous-time systems, it talks a bit about chaos, and it gives a brief description of Poincaré/return maps that are often used in the study of approximately periodic systems.

My point is that a discrete-time system really cannot be interpreted within a field-theoretic framework. A "phase portrait" captures both position and momentum of a continuous-time system described by an ordinary differential equation. These momentum variables setup the "field" that gives structure to the phase portrait. In a discrete-time system, we don't have the same kind of momentum.

     For a continuous-time system, I can plot a point at an individual position, and I can also then draw a vector pointing away from that point representing the velocity at that instant of time. It is these velocity vectors that are put together to make a phase portrait of the system.

     For a discrete-time system, there is no point-based velocity. In order for me to calculate the approximate "velocity" at a point, I need to know the position at the next point. Then I can draw a line between those two points and approximate the "velocity" of the system as going from the first point to the second point. However, if I have to know the next point in future anyway, it's more useful to just draw the second point.

     Now, some discrete-time systems have a more predictable structure. For example, if you have a linear time-invariant discrete-time system like:

x[k+1] = M*x[k]

then the algebraic structure of the "M" matrix gives us insight into how trajectories will evolve. So for these special cases, it is possible to draw a kind of "phase portrait" for the discrete-time system. However, this is primarily because such a discrete-time system can be viewed as a sampled version of a continuous-time linear time-invariant system which does have a phase portrait.

     So, for an arbitrary discrete-time system, the best thing you can do is explore trajectories from different initial conditions. Gradually, as you explore the space more and more, you may find boundaries of attractors (possibly strange attractors). A complication with discrete-time systems is that the "next" point may be very far from the "previous" point. Take, for example:

x[k+1] = -1.1*x[k]

If you start at x[0]=1, the trajectory will bounce from point clustered above 1 to points clustered below -1. In a continuous-time system, you might expect to see initial conditions above 1 stay near 1 and initial conditions below -1 stay near -1. That is, in a continuous-time system, you wouldn't imagine trajectories could cross x=0 (which is an equilibrium/fixed-point of this system). However, the discrete-time system can jump wildly from point to point.

     Take, for example, the Henon map you mention. Wolfram's Mathworld has some nice plots:

http://mathworld.wolfram.com/HenonMap.html

The first pair of side-by-side plots are colored "according to the number of iterations required to escape". That is, the plots were generated by starting at several initial conditions and recording the resulting trajectories. Each "iteration" gives the next point from the previous point. For a while, a trajectory will stay around its initial condition. Eventually, it will escape and move away from the region. The regions are colored based on how many iterations (i.e., how many calculations after the initial condition) it took for the trajectory to leave the region.

     The second pair of side-by-side plots show a SINGLE trajectory started at x[0]=0 and y[0]=0. Each point was recorded and gradually a pattern emerged. Notice how in the left plot the two regions appear to be disconnected. If you saw a phase portrait that looked like this in a continuous-time system, you would conclude that initial conditions within one region would not be able to join the other region for this set of parameters. However, this plot was generated from a SINGLE initial condition. So the plot jumps from points in the top left to points in the bottom right and back.

     So that's how you can explore something like the "phase space" for discrete-time systems. You can probe it with different initial conditions. For chaotic systems, you have to be very careful you don't accidentally jump over an interesting region of initial conditions that may have qualitatively different trajectories that follow from them.

     As an aside, I guess it's also worth mentioning that many popular discrete-time chaotic maps are actually Poincaré maps of continuous-time dynamical systems. Poincaré maps have other names, including "return maps." Consider, for example, the planets as they orbit the sun. The actual orbits of the planets in three dimensions looks like a tangled mess when you consider their histories over several cycles around the sun because each orbit is slightly different than the previous orbit (i.e., they aren't entirely planar). However, if you insert a plane perpendicular to their orbits at a single location, each planet pierces the plane at one point every cycle. The resulting shapes that are poked out of that cross section reveal structure in the orbits.

     I hope that helps! --
     Ted

Tuesday, October 09, 2012

natbib-like frontend for chicago-style macros

For some reason, the ACM has baked in chicago-like citation macros into the ACM SIG proceedings LaTeX templates instead of using the far superior natbib that literally everyone else on the planet uses. I'm much more accustomed to typing \citet as opposed to \shortciteN, \citeauthor as opposed to \shortciteANP, etc. So I decided to add this little translation table to my preamble:
\def\citet{\shortciteN}
\def\Citet{\shortciteN}
\def\citeauthor{\shortciteANP}
\def\Citeauthor{\shortciteANP}

\let\chicagociteyear\citeyear
\def\citeyearpar{\chicagociteyear}
\def\citeyear{\citeyearNP}

\def\citep{\shortcite}
Unfortunately, that means I can't use any of the optional arguments to \citep or \citet. If you are just looking for a translation table so you can use the chicago-style macros directly, try this:
natbibchicago
\citet or \Citet\shortciteN
\citeauthor or \Citeauthor\shortciteANP
\citeyear\citeyearNP
\citeyearpar\citeyear
\citep\shortcite
It's crazy how chicago-style macros sometimes use "N" to mean "noun" and other times to be the first half of "NP" being "no parentheses."